PRE2015 3 Groep1 considerations: Difference between revisions

From Control Systems Technology Group
Jump to navigation Jump to search
No edit summary
No edit summary
 
(17 intermediate revisions by the same user not shown)
Line 1: Line 1:
== Safety ==
== Safety and Violence ==
Drones require several limits to reduce the chance for risks and accidents. Every drone will be facing technological problems which are inevitable, but some interventions may reduce their occurance or their impact.


== Privacy ==
The more complicated a drone will be the more (technological) problems are given the possibility to intervene. For example, the more autonomously the drone is, the more ethical and technological issues are to be resolved. Therefore the more complex the drone and its actions, the greater the decrease of it’s reliability. This is why it is usefull to keep a drone as simple as possible.
 
Reducement of navigation errors or obstacle collision is reached by sufficiend data feeds of the surrounding of the drone. The better the information is about the surrounding the better the drone will function. It is also favourable to have a more or less continuous data feed. This reduces the possibility of missing information or obstacles and therefore reduces collisions. Another navigation improving property is the amount of training data or options a drone is given. By having more options in certain situations , the drone is more likely to be able to choose the right choice. Of course there must not be that many options for the drone to be slow in decision making, because this would have a rather negative effect on the decision making.
 
In our concept the drone communicates with the safety guard , and this safety guard has the final decision responsibility. This has consequences for the drones actions. Therefore the communication of the drone with other drones or the surveillance center Is very important. Airwaves or electronic communications can interfere in the drones communication and prevent the message to arrive at the surveillance center or back at the drone. Because those messages have a huge impact on the drones actions, it is important to secure the communication to ensure a safe surrounding for the intruders which are probably close to the drone at that moment. Cheap drones are mostly quite easy to hack. Therefore a certain amount of costs must probably be made to secure the drone to the right level.
 
Any collision with a drone easily has a big impact. Not only the kinetic energy the drone exerts is of importance, the propellors too can cause serious trauma’s or injury’s. The damage by the drone is less if the drone is smaller if the drone itself is smaller too. This requires a drone as small as possible for our application. On the other hand, a drone too small or light is too easily affected by external forces like the wind or lightning. A good compromise must be found between the two of these.
 
Besides the physical damage the drone itself can cause, it can distract or interfere with other vehicles which will cause accidents. In this project we won’t include intentionally induced harm, the drone is not made to fight the intruder. Solely to slow it down or capture its face in a picture.
 
Furthermore the reaction of the drone in certain unexpected situations is very important. By choosing a drone which for example can still land safely when one of its propellors has a ddefect we can ensure more safe situations. Even so it has to autmatically react on low battery signals and be able to reach the docking station before the battery is fully empty
 
Weather conditions are very difficult for drones. The wind can displace the drone, and it must be able to find its way back to it’s surrounding.This wind can also make the drone suspectible for hitting a person if it’s too close and the wind catches on. the drone needs to take the wind in account in staying in a safe distance from obstacles and especially persons
 
== Privacy <ref>Mr.ir. A. Engelfriet ''Cameratoezicht, filmen en fotograferen van mensen (januari 2014)''. Available from: <http://www.iusmentis.com/maatschappij/privacy/filmen-cameratoezicht/#openbareruimte> (23 february 2016).</ref> ==
Privacy is a very important subject in the use of security drones because these drones can film and record everything and everybody they see. Because of this, is it necessary to find out more about the rules and laws around filming in public and private area.
 
===Public areas===
First of all, the laws says that any filming or photographing of a person with a mounted camera is forbidden, unless this is clearly announced beforehand.
This means that taking a photo or filming with for example a smartphone without announcing, is allowed. It is also allowed to film with a fixed camera as long as you make sure nobody that is being recorded can be recognized. These rule apply in any public domain. This is not just the street. This also holds for restaurants for example. Despite the fact that the restaurant is private property, anyone can come in so it is a public domain (it is also irrelevant whether it you have to pay in order to enter or not). You do not necessarily have to inform people with a written sign. It is also sufficient if, for example, the camera is immediately visible for the person entering the area.  Some stores hang a monitor next to the entrance so people can see what the camera sees. This is also sufficient.
 
To summarize this: Filming or photographing people in public domain is only allowed if everyone in that specific area knows (in some way) about the fact that they could be filmed (or photographed).
 
===Private areas===
The law says that recording (or photographing) in private area is forbidden if this is not clearly announced beforehand. There is no difference in recording with a fixed camera and recording with a ‘’pocket camera’’ (smartphone, digital camera). This does not hold for recording in your own house, garden or office.
 
The prohibition on recording in private areas is based on violating the privacy of the recorded persons in these areas. However, this does not hold for a person that has illegally entered that area. So using a camera to record an burglar is allowed, but this does not mean that you can record anyone because they might be a burglar.
 
The laws says that recording is allowed (in principle) if this is clearly announced beforehand. However, it could still be allowed to film without clearly announcing this, but only when there is a strong reason to do so.
 
===Use for evidence===
It is allowed for civilians or companies to hand over footage to the police if there are illegal acts on this footage. Even if this footage is obtained illegally. The footage can be used as evidence because it was a civilian who was violating the law. This would be different if the police would be involved in acquiring the footage.
This does not mean that the civilian who made the recordings goes unpunished. Even if the footage can be used as evidence, the filmmaker still risks a conviction.
 
===Relation to security drones===
 
Security drones use cameras to film what is happening around them. The big question is: What privacy problems occur when using a security drone? The law contains little information about using drones with cameras for security. There are some articles available about privacy regarding the government using surveillance drones. But in our situation, the drones are owned by private companies. So, for now, we will try to use existing laws for regular cameras for implementing security drones.
Because the cameras that security drones use are not in the hand of a person, we consider these cameras to be fixed cameras. This means that it is always obligated to inform people that are near these drones about the fact that they could be recorded. Implementing this indoors can easily be done by placing signs at each entrance of the secured building. Implementing this for a private outdoor area can be a bit harder. But it would still be possible to place signs around the area that warn you that it is forbidden to enter the area as well as that are drones with cameras in the area. These signs can automatically warn people that are close to the area, but still in a public area, about the cameras, which should cover the problem of filming with a “fixed” camera in a public area.
 
The last problem that needs to be solved, has to do with following an escaping suspect. When a drone is following a suspect, there is a big possibility that the suspect will leave the (public) area where these signs are placed. This would mean that the drone is filming illegally in a public domain. This can cause regulatory problems. However, one could argue that it is authorized to film in this case because there is a strong reason to do so.


== Decision making ==
== Decision making ==
In order to establish which decisions are made by the security guard and which by the drone itself, the level of autonomy has to be concluded first. The following subjects are separated: ''Navigating'', ''React to suspects'' and ''Following a person''
===Navigating===
There are already a lot of drones that can fly around fully autonomous. The most used design is the reactive autonomy which is based on the neural systems of insects <ref>D. Floreano & R. J. Wood ''Science, technology and the future of small autonomous drones''</ref>. The lightweight sensors of the drone create a vision of the world based on the relative distance to their surroundings. This is called the optic flow. Using algorithms processing the optic flow, drones can tackle a lot of flight problems. It is possible to regulate capabilities like flight speed, wall following, odometry estimation, altitude regulation and even obstacle avoidance. The drones will be programmed to follow a certain path depending on the situation. And when an obstacle is in their way they use the optic flow to find another path. It can be concluded that the drone can make the ‘decisions’ in respect to navigation all by itself. This is much cheaper compared to a situation where human need to control the drone navigating.


== Violence ==
===React to suspects===
The drone is equipped with motion sensors that measure the relative speed of the ground and other surrounding objects. This way it can detect any movement in the area. When a moment is detected the drone must make a analyses what were the movement came from and a decision what to do next. In our current design the drone will not be able to give a good analyses of the situation because it will not know the difference between a human and other causes of movement. It could be possible to give to drone more tools to examine the situation such as heat sensors but this will make the drone heavier, more complicated and still not trustworthy. For instance it would not know the difference between a human and an animal. The drone can also distinguish color on the footage it is making, but humans have no standard color of clothes, skin and hair. That is why there must be human involvement in this step. The drone will send a signal to the security guard that some movement is detected and add live visuals via its camera. This way the guard can inspect the scene and decide what action suits best. Possible decisions are: let the drone continue on its route, let the drone follow the person or simply call the police.
 
===Following a person===
The next question that raises is how the drones can follow a person while they cannot detect one with high certainty. The key answer to question is combining information. When the security guard signals the drone that the detected movement is a person and that the drone must follow it. The drone saves the color as was detected from the source of the movement. Then the color is matched to the person. Now the drone can follow the color by staying on a certain distance of the object. This technique is used in the project drone4u <ref>T. Tanzi, L. Apvrille and J.-L. Dugelay ''Drones4u project  (16 may 2015)''</ref>  where the drones follow person in different circumstances. The drones can deal with light variations, turning and running of the person, height alternations and even short loss of sight. The biggest problem with this technique is the localization of the person when he/she is camouflaged because it relies on color contrast. It seems that the drone can follow a person by itself, but the system will not work perfectly in every situation. In this project the choice is made that the drone will try follow the suspects autonomously. Another way to identify a person is by using a Histogram of Gradients (HOG) <ref>A. Ludovic ''Autonomous drones for assisting rescue services (23 aug 2016)''</ref>. Also based on color the persons are spotted by so called Regions of Interest (ROL). The difference is that a particle filter is used to predict where the person will go next based on the prior information. This way the drone will be able to follow a person.


== Applicability indoors ==
== Applicability indoors ==
The applicability of drones indoors is not so trivial as for outside. A lot of design considerations - for the drone as well as the building - must be made, to make sure the drone is able to operate as it should. Size, maneuverability, knowledge of the building (and the patrolling paths) are the most important considerations
===Size===
The first restriction that has to be looked at is the size of the drones. Generally, buildings consist of narrow passageways and small rooms. When a drone is used as a security guard and stationary camera replacement, it must be able to see at least the things a camera can see, and must be able to go at least anywhere a security guard can go. This means the drone has to be small enough to navigate in tight spaces. Though if the building to be protected is a museum just like in the scenario, the size of the drone is less important. This, since a museum has the exhibits often placed in spacious rooms, where the drones can roam freely.
Looking at the size of the buildings, it is desired that the drones are as small as possible. However, there are other aspects influencing the drone’s size. Smaller drones are generally more fragile, because they lack the protection of the propellers for example. On the other hand, smaller drones are less likely to be hit by the perpetrators.
Besides this last remark, the drone also has to carry the needed equipment for detecting perpetrators and the environment. Of course, it does not have to carry the big static surveillance cameras that are installed in buildings right now, since small digital cameras also exist, which can already be found in smartphones. The quality of these type of cameras is good enough and will not make the drone large in size.
===Maneuverability===
Since the drone must be able to move in small spaces and possibly chase perpetrators, it is important that the drone is very maneuverable. The maneuverability of the drone is related to a couple of things, of which size is the first. Ideally, to increase the maneuverability of the drone, it has to be as small as possible. But, as mentioned earlier, this could lead to constraints regarding the surveillance equipment.
The propulsion of the drone is the second important aspect in the maneuverability. As touched upon earlier in the state of the art, there already exist a couple of ways for drones to move around. Table 1 shows the different ways of advancing together with the advantages and disadvantages.
{| class="wikitable"
|-
! Propulsion
! Advantages
! Disadvantages
|-
| Floaters
| Can hover, lightweight, not expensive, small, long flight range
| Low speed, fragile
|-
| Planes
| Long range, fast, sturdy
| Cannot hover, heavy, very expensive, large
|-
| Quad-/ octocopters
| Can hover, small, high speed, fast flight corrections possible
| Noisy, low flight range, fragile (if propellers are not protected)
|-
| Butterfly
| Lightweight, not expensive
| Slow, cannot hover, fragile, camera footage is not steady because of vibration
|}
===Knowledge===
Knowledge of the building is also an important aspect of the indoor drones, since they have to know where they are and if any changes in the environment occurred. Together with GPS, a map of the building can be used. However, since GPS is only accurate up to a few meters, this is not the best way to determine position of the drone.
An alternative could be an environment recognition system, which uses the footage from the camera and compares it to already taken photos of the building. These photographs are then linked to certain positions in the building, just like how the Layer-app works. In this way, the drone can determine its position
It is also possible to make use of beacons in the building. These beacons can send out a signal which is distinct for a certain part of the building. In this way, with the signals from the beacons, the drone knows its position in the building. The disadvantage of this system is that it requires modifications to the building. This makes it less universal and more difficult to implement.
== References ==
<references />

Latest revision as of 13:18, 2 March 2016

Safety and Violence

Drones require several limits to reduce the chance for risks and accidents. Every drone will be facing technological problems which are inevitable, but some interventions may reduce their occurance or their impact.

The more complicated a drone will be the more (technological) problems are given the possibility to intervene. For example, the more autonomously the drone is, the more ethical and technological issues are to be resolved. Therefore the more complex the drone and its actions, the greater the decrease of it’s reliability. This is why it is usefull to keep a drone as simple as possible.

Reducement of navigation errors or obstacle collision is reached by sufficiend data feeds of the surrounding of the drone. The better the information is about the surrounding the better the drone will function. It is also favourable to have a more or less continuous data feed. This reduces the possibility of missing information or obstacles and therefore reduces collisions. Another navigation improving property is the amount of training data or options a drone is given. By having more options in certain situations , the drone is more likely to be able to choose the right choice. Of course there must not be that many options for the drone to be slow in decision making, because this would have a rather negative effect on the decision making.

In our concept the drone communicates with the safety guard , and this safety guard has the final decision responsibility. This has consequences for the drones actions. Therefore the communication of the drone with other drones or the surveillance center Is very important. Airwaves or electronic communications can interfere in the drones communication and prevent the message to arrive at the surveillance center or back at the drone. Because those messages have a huge impact on the drones actions, it is important to secure the communication to ensure a safe surrounding for the intruders which are probably close to the drone at that moment. Cheap drones are mostly quite easy to hack. Therefore a certain amount of costs must probably be made to secure the drone to the right level.

Any collision with a drone easily has a big impact. Not only the kinetic energy the drone exerts is of importance, the propellors too can cause serious trauma’s or injury’s. The damage by the drone is less if the drone is smaller if the drone itself is smaller too. This requires a drone as small as possible for our application. On the other hand, a drone too small or light is too easily affected by external forces like the wind or lightning. A good compromise must be found between the two of these.

Besides the physical damage the drone itself can cause, it can distract or interfere with other vehicles which will cause accidents. In this project we won’t include intentionally induced harm, the drone is not made to fight the intruder. Solely to slow it down or capture its face in a picture.

Furthermore the reaction of the drone in certain unexpected situations is very important. By choosing a drone which for example can still land safely when one of its propellors has a ddefect we can ensure more safe situations. Even so it has to autmatically react on low battery signals and be able to reach the docking station before the battery is fully empty

Weather conditions are very difficult for drones. The wind can displace the drone, and it must be able to find its way back to it’s surrounding.This wind can also make the drone suspectible for hitting a person if it’s too close and the wind catches on. the drone needs to take the wind in account in staying in a safe distance from obstacles and especially persons

Privacy [1]

Privacy is a very important subject in the use of security drones because these drones can film and record everything and everybody they see. Because of this, is it necessary to find out more about the rules and laws around filming in public and private area.

Public areas

First of all, the laws says that any filming or photographing of a person with a mounted camera is forbidden, unless this is clearly announced beforehand.

This means that taking a photo or filming with for example a smartphone without announcing, is allowed. It is also allowed to film with a fixed camera as long as you make sure nobody that is being recorded can be recognized. These rule apply in any public domain. This is not just the street. This also holds for restaurants for example. Despite the fact that the restaurant is private property, anyone can come in so it is a public domain (it is also irrelevant whether it you have to pay in order to enter or not). You do not necessarily have to inform people with a written sign. It is also sufficient if, for example, the camera is immediately visible for the person entering the area. Some stores hang a monitor next to the entrance so people can see what the camera sees. This is also sufficient.

To summarize this: Filming or photographing people in public domain is only allowed if everyone in that specific area knows (in some way) about the fact that they could be filmed (or photographed).

Private areas

The law says that recording (or photographing) in private area is forbidden if this is not clearly announced beforehand. There is no difference in recording with a fixed camera and recording with a ‘’pocket camera’’ (smartphone, digital camera). This does not hold for recording in your own house, garden or office.

The prohibition on recording in private areas is based on violating the privacy of the recorded persons in these areas. However, this does not hold for a person that has illegally entered that area. So using a camera to record an burglar is allowed, but this does not mean that you can record anyone because they might be a burglar.

The laws says that recording is allowed (in principle) if this is clearly announced beforehand. However, it could still be allowed to film without clearly announcing this, but only when there is a strong reason to do so.

Use for evidence

It is allowed for civilians or companies to hand over footage to the police if there are illegal acts on this footage. Even if this footage is obtained illegally. The footage can be used as evidence because it was a civilian who was violating the law. This would be different if the police would be involved in acquiring the footage. This does not mean that the civilian who made the recordings goes unpunished. Even if the footage can be used as evidence, the filmmaker still risks a conviction.

Relation to security drones

Security drones use cameras to film what is happening around them. The big question is: What privacy problems occur when using a security drone? The law contains little information about using drones with cameras for security. There are some articles available about privacy regarding the government using surveillance drones. But in our situation, the drones are owned by private companies. So, for now, we will try to use existing laws for regular cameras for implementing security drones. Because the cameras that security drones use are not in the hand of a person, we consider these cameras to be fixed cameras. This means that it is always obligated to inform people that are near these drones about the fact that they could be recorded. Implementing this indoors can easily be done by placing signs at each entrance of the secured building. Implementing this for a private outdoor area can be a bit harder. But it would still be possible to place signs around the area that warn you that it is forbidden to enter the area as well as that are drones with cameras in the area. These signs can automatically warn people that are close to the area, but still in a public area, about the cameras, which should cover the problem of filming with a “fixed” camera in a public area.

The last problem that needs to be solved, has to do with following an escaping suspect. When a drone is following a suspect, there is a big possibility that the suspect will leave the (public) area where these signs are placed. This would mean that the drone is filming illegally in a public domain. This can cause regulatory problems. However, one could argue that it is authorized to film in this case because there is a strong reason to do so.

Decision making

In order to establish which decisions are made by the security guard and which by the drone itself, the level of autonomy has to be concluded first. The following subjects are separated: Navigating, React to suspects and Following a person

Navigating

There are already a lot of drones that can fly around fully autonomous. The most used design is the reactive autonomy which is based on the neural systems of insects [2]. The lightweight sensors of the drone create a vision of the world based on the relative distance to their surroundings. This is called the optic flow. Using algorithms processing the optic flow, drones can tackle a lot of flight problems. It is possible to regulate capabilities like flight speed, wall following, odometry estimation, altitude regulation and even obstacle avoidance. The drones will be programmed to follow a certain path depending on the situation. And when an obstacle is in their way they use the optic flow to find another path. It can be concluded that the drone can make the ‘decisions’ in respect to navigation all by itself. This is much cheaper compared to a situation where human need to control the drone navigating.

React to suspects

The drone is equipped with motion sensors that measure the relative speed of the ground and other surrounding objects. This way it can detect any movement in the area. When a moment is detected the drone must make a analyses what were the movement came from and a decision what to do next. In our current design the drone will not be able to give a good analyses of the situation because it will not know the difference between a human and other causes of movement. It could be possible to give to drone more tools to examine the situation such as heat sensors but this will make the drone heavier, more complicated and still not trustworthy. For instance it would not know the difference between a human and an animal. The drone can also distinguish color on the footage it is making, but humans have no standard color of clothes, skin and hair. That is why there must be human involvement in this step. The drone will send a signal to the security guard that some movement is detected and add live visuals via its camera. This way the guard can inspect the scene and decide what action suits best. Possible decisions are: let the drone continue on its route, let the drone follow the person or simply call the police.

Following a person

The next question that raises is how the drones can follow a person while they cannot detect one with high certainty. The key answer to question is combining information. When the security guard signals the drone that the detected movement is a person and that the drone must follow it. The drone saves the color as was detected from the source of the movement. Then the color is matched to the person. Now the drone can follow the color by staying on a certain distance of the object. This technique is used in the project drone4u [3] where the drones follow person in different circumstances. The drones can deal with light variations, turning and running of the person, height alternations and even short loss of sight. The biggest problem with this technique is the localization of the person when he/she is camouflaged because it relies on color contrast. It seems that the drone can follow a person by itself, but the system will not work perfectly in every situation. In this project the choice is made that the drone will try follow the suspects autonomously. Another way to identify a person is by using a Histogram of Gradients (HOG) [4]. Also based on color the persons are spotted by so called Regions of Interest (ROL). The difference is that a particle filter is used to predict where the person will go next based on the prior information. This way the drone will be able to follow a person.

Applicability indoors

The applicability of drones indoors is not so trivial as for outside. A lot of design considerations - for the drone as well as the building - must be made, to make sure the drone is able to operate as it should. Size, maneuverability, knowledge of the building (and the patrolling paths) are the most important considerations

Size

The first restriction that has to be looked at is the size of the drones. Generally, buildings consist of narrow passageways and small rooms. When a drone is used as a security guard and stationary camera replacement, it must be able to see at least the things a camera can see, and must be able to go at least anywhere a security guard can go. This means the drone has to be small enough to navigate in tight spaces. Though if the building to be protected is a museum just like in the scenario, the size of the drone is less important. This, since a museum has the exhibits often placed in spacious rooms, where the drones can roam freely.

Looking at the size of the buildings, it is desired that the drones are as small as possible. However, there are other aspects influencing the drone’s size. Smaller drones are generally more fragile, because they lack the protection of the propellers for example. On the other hand, smaller drones are less likely to be hit by the perpetrators.

Besides this last remark, the drone also has to carry the needed equipment for detecting perpetrators and the environment. Of course, it does not have to carry the big static surveillance cameras that are installed in buildings right now, since small digital cameras also exist, which can already be found in smartphones. The quality of these type of cameras is good enough and will not make the drone large in size.

Maneuverability

Since the drone must be able to move in small spaces and possibly chase perpetrators, it is important that the drone is very maneuverable. The maneuverability of the drone is related to a couple of things, of which size is the first. Ideally, to increase the maneuverability of the drone, it has to be as small as possible. But, as mentioned earlier, this could lead to constraints regarding the surveillance equipment.

The propulsion of the drone is the second important aspect in the maneuverability. As touched upon earlier in the state of the art, there already exist a couple of ways for drones to move around. Table 1 shows the different ways of advancing together with the advantages and disadvantages.

Propulsion Advantages Disadvantages
Floaters Can hover, lightweight, not expensive, small, long flight range Low speed, fragile
Planes Long range, fast, sturdy Cannot hover, heavy, very expensive, large
Quad-/ octocopters Can hover, small, high speed, fast flight corrections possible Noisy, low flight range, fragile (if propellers are not protected)
Butterfly Lightweight, not expensive Slow, cannot hover, fragile, camera footage is not steady because of vibration

Knowledge

Knowledge of the building is also an important aspect of the indoor drones, since they have to know where they are and if any changes in the environment occurred. Together with GPS, a map of the building can be used. However, since GPS is only accurate up to a few meters, this is not the best way to determine position of the drone.

An alternative could be an environment recognition system, which uses the footage from the camera and compares it to already taken photos of the building. These photographs are then linked to certain positions in the building, just like how the Layer-app works. In this way, the drone can determine its position

It is also possible to make use of beacons in the building. These beacons can send out a signal which is distinct for a certain part of the building. In this way, with the signals from the beacons, the drone knows its position in the building. The disadvantage of this system is that it requires modifications to the building. This makes it less universal and more difficult to implement.

References

  1. Mr.ir. A. Engelfriet Cameratoezicht, filmen en fotograferen van mensen (januari 2014). Available from: <http://www.iusmentis.com/maatschappij/privacy/filmen-cameratoezicht/#openbareruimte> (23 february 2016).
  2. D. Floreano & R. J. Wood Science, technology and the future of small autonomous drones
  3. T. Tanzi, L. Apvrille and J.-L. Dugelay Drones4u project (16 may 2015)
  4. A. Ludovic Autonomous drones for assisting rescue services (23 aug 2016)