PRE2019 3 Group9

Research robot for the burrows of prairie dogs

Group members

Name	Study	Student Number
Nick Reniers	Technische Wiskunde	1258362
Jankatiri Boon	Werktuigbouwkunde	1003254
Milan Hutten	Software Science	0997241
Mendel van der Vleuten	Technische Wiskunde	1262564
Ferenc Sterkens	Werktuigbouwkunde	1022090

Introduction

A lot of research has been done into flora and fauna all over the world, but our knowledge of subterranean species remains relatively limited. This is in large part due to our inability to study these animals without destroying their burrows and scaring them away. Most current techniques require destruction of subterranean burrows or even killing the animals. This makes certien fields of study such as studying the behaviour of species very difficult.

Problem statement

Animal researchers are unable to effectively gather data of subterranean species without destroying their burrows or tunnel systems.

• Little knowledge about subterranean species
• Hard to reach, as a result burrows often destroyed and animals scared away or killed
• Makes certain field of study hard, such as study of animal behaviour difficult
• One such species of which their behaviour underground remains largely unknown is the black-tailed prairie dog
• The use of certain of their underground tunnels and rooms remains a mystery
• goal: increasing knowledge about prairie dogs

Users

Researchers of subterranean animals or underground tunnel systems

Users	Society	Enterprises
Researchers	Towns closeby	Robot manufacturers
Engineers		Universities
Animal caretakers		Zoo's

User Needs

The following list of user needs does not focus on any particular type of researcher or animal species yet and thus requires modification

• Observe animals within their burrows
• Visualize burrows
• Keep track of population growth
• Safely (both for the researcher and the animal) extract animal from their burrow

Animal Needs

The following list of user needs does not focus on any particular animal species yet and thus requires modification

• Be disturbed as little as possible (May require more elaborate description)

Biological study fields

This list is composed purely based on the short descriptions of the studies.

• Ethology: the study of the behaviour of animals. Requires ability to observe individuals.
• Entomolgy, Herpetology, Ichtyology, Mammalogy and Ornithology: the studies of insects, reptiles and amphibians, fish, mammals and birds respectively. Purely based on the species we focus on.
• Biogeograph: the study of the distribution of species spatially and temporally. Requires ability to observe individuals and count populations.
• Biomechanics: the study of the mechanics of living beings. Requires ability to observe individuals. Specifically interesting for mechanics that do not or rarely occur above ground.
• Chronobiology: the study of periodic events in living systems. Requires ability to observe individuals
• Conservation biology: the study of the preservation, protection, or restoration of the natural environment, natural ecosystems, vegetation, and wildlife. Requires ability to count populations.
• Ecology: the study of the interactions of living organisms with one another and with the non-living elements of their environment. Requires ability to observe individuals.
• Sociobiology: the study of social behavior in terms of evolution. Requires ability to observe individuals.

Requirements

The robot should be able to:

• Safely and autonomously navigate the specified underground systems
• Map these systems adequately
• Be able to return to the user after completing its tasks

Approach

We approach the problem in a very practical manner, we opt to create a robot that autonomously investigates underground tunnels and maps them. We first make a selection of subterranean animals for which we can map their corresponding burrows, and then research details of these animals and underground systems as to prepare a robot that can safely navigate them

Objectives and milestones

• Make a selection of animals for which it is feasible to construct a robot that navigates their burrows
• Research the animals specified in the first milestone and their corresponding underground systems
• Make a construction plan for a robot that could navigate said tunnels adequately
• Prepare software for path finding in burrowss
• Prepare software for mapping the underground systems
• Construct the robot
• Validate the workings of the robot and summarize our findings

Task division

State-of-the-art

Interview Candidates

prof.dr.ir F (Frank) van Langevelde has only 3 publications related to soil fauna. In these studies, they mainly refer with "soil fauna" to smaller critters such as nematodes, collembola (springtails) and other small arthropods inhabiting the soil. Most of which either do not create burrows but just tunnels through moving around in the soil. Their sampling of soil fauna can be oversimplified as extracting samples from the natural environment, extracting the desired type of animal from those samples, for example extracting collembola using tullgren funnels, and doing the required measurements. Because most species do not create burrows and because of the effectiveness of the current sampling strategy, the use of robots in this scenario does not seem to be a viable option.

• Van Langevelde, F., V. Comor, S. de Bie, H.H.T. Prins and M.P. Thakur (2020) Disturbance regulates the density – body mass relationship of soil fauna. Ecological Applications 30:e02019
• Thakur, M.P., M.P. Berg, N. Eisenhauer and F. van Langevelde (2014) Disturbance–diversity relationships for soil fauna are explained by faunal community biomass in a salt marsh. Soil Biology & Biochemistry 78:30–37
• Comor, V., M.P. Thakur, M.P. Berg, S. de Bie, H.H.T. Prins and F. van Langevelde (2014) Productivity affects the density – body mass relationship of soil fauna communities. Soil Biology & Biochemistry 72:203–211

dr.ir. AR (Anouschka) Hof has mostly done studies into the decline of the western hedgehogs pupulations in europe, and mainly in Great Britain, when it comes to soil fauna. This makes her an interesting candidate, though there are two things to consider. First, she seems to mostly get data via questionaires rather than field study. Just like with the sampling of the soil fauna done by Langevelde, the questionaire might be more effective then using robots. Secondly, a hedgehogs den is not very complex, and not even always underground, making pathfinding not a big priority. If a robot would be used, a remote controlled one would probably easily suffice.

• The impact of grassy field margins on macro-invertebrate abundance in adjacent arable fields
• A study of the current status of the hedgehog (Erinaceus europaeus), and its decline in Great Britain since 1960
• The value of green-spaces in built-up areas for western hedgehogs
• Quantifying the long-term decline of the West European hedgehog in England by subsampling citizen-science datasets
• Factors affecting hedgehog presence on farmland as assessed by a questionnaire survey
• European terrestrial gastropod distribution. How may climate change affect their diversity and current distribution
• Investigating the role of the eurasian badger (Meles meles) in the nationwide distribution of the western european hedgehog (Erinaceus europaeus) in England
• Local variations in small scale movements of hedgehogs in rural areas.
• Egels in de problemen?

Y (Yorick) Liefting BSc is not a biological scientist himself, but he experiments with new technology and develops tools to support research.

Path finding

Given that we want to map underground tunnels of a specific species of subterranean animals, we need a systematic and preferably mathematically adequate way to represent different important aspects of the tunnels. The most straightforward way to do this is by the usage of the mathematical objects that are graphs. A graph G is a tuple G=(V,E) consisting of a vertex set V and an edge set E. Here, the edge set E contains unordered pairs of vertices between which an edge is present. There is also a variant of graphs in which the vertex pairs are ordered, but for the purpose of tunnel mapping, this is inconvenient. We can also consider the concept of a multigraph, in which multiple edges are possible between the same two vertices, this may be the case for some certain species of animals, but this demands further research. For now, we will assume simple undirected graphs as representatives for the tunnels of our species of animals.

For the purpose of mapping an underground tunnel system in general, one can deploy a multitude of graph algorithms. The most obvious algorithms for this purpose are depth first search and breadth first search, which greedily traverse a graph (representing the underground system of the animal). We can also look into finding specific points of interest depending on the animals we want to investigate. For example, one may be particularly interested in finding a nest in a mole underground system. For such purposes, specialized search algorithms that take into account the known properties of such points of interest, may be optimal.

Given that we have found the graph corresponding to the subterranean structure we want to investigate, we can look at some interesting properties of the graph. Examples would be the minimum spanning tree of the graph or the (non-)existence of cycles and connectedness.

When talking about path finding in robots it is also necessary to think of ways for the robot to follow a path. Considering we will be working underground and partially above ground in places made constructed by animals, it would be a good idea to look into how these animals move or to inspire the movement of the robot on biological movement. In the paper "Biologically Inspired Locomotion Strategies: Novel Ground Mobile Robots at RoMeLa" a few designs are discussed on moving forward on land. The first robot is a concept that is closely related to tracks. This design is inspired on single celled organisms such as amoeba. An elongated torus continuously rolls such that there is a system similar to 360 degree track system. This could be very useful for our goals, but difficulties will lie in gathering and collecting data. The second design involves a robot with 3 legs, This design makes use of a swinging third legs, so it needs a lot of space. This is not something which would be useful for underground or small above ground burrows/nests. The robot MARS utilizes six axi-symmetrically arranged limbs, this robot is well adapted for crawling over uneven terrain, which is what we are interested in. But it does not perform well in tunnel systems. It also uses gecko-like dry adhesive to stick to surfaces, but as we are not dealing zero gravity this will most likely not be necessary. So depending on what our final goal is this robot may or may not be of interest. The IMPASS robot has wheels like structures to move forward, it has 6 spokes with small foot like platforms so it can roll. The thing that makes this robot stand out is that the spokes can change in length allowing for better traversing of rough terrain and more stability. Again this robot seems very adept at walking on ground, but less so underground. Finally a humanoid robot is presented but this is very impractical for our goals.

Prairie dogs

Prairie dogs are herbivorous rodents that create interesting burrows. They are native to the grasslands of North America. There are five species of prairie dogs: black-tailed, white-tailed, Gunnison's , Utah, and Mexican prairie dogs.

Black-tailed prairie dogs

Description

Usually, black-tailed prairie dogs have a body that is 36 to 43 centimeters tall and weighs roughly 700 to 1500 grams, where the males tend to be heavier than the females. Their bodies are also very compact, and they have black long claws that are used for digging. The black tailed prairie dogs are usually colored a pale tone of brown, with a black spots on their tail. These points of information may be useful for letting our robot recognize the animals when exploring their burrows, alternatively we may use heat sensors to detect living beings.

Social life and habits

Black-tailed prairie dogs live in prairies of western North America. An image of their geographical distribution is shown below:

The species is also very colonial, colonies of the black-tailed prairie dogs may contain thousands of members, their territory stretching out kilometers in all directions. Colonies are further subdivided in a couple of wards that are based around topographic features such as hills. These ward themselves can then be subdivided into coteries. These coteries consist of groups of one or two men with a number of women and their offspring.

Black-tailed prairie dogs are also diurnal, meaning that they are typically active during daytime. As such, their foraging tends to happen from dawn to dusk. During winter, they do not hibernate and continue to forage. At night however, they go in a state of torpor which is similar to hibernation but shorter and involuntary.

Burrows

Burrows are underground tunnels that have at least one opening at the surface. Prairie dog burrows are essential to defend the species against predators and weather circumstances. The diameter of such a burrow is usually around 10-30cm at the surface and a little bit more narrow underground. The depth of these burrows is generally around 2-3 meters with a length of 5 to 10 meters, but they can range to 33 meters long and 10 meters deep. Burrows can have one of multiple functions, spending the night, bringing up offspring or refuge from weather circumstances and predators. This may then explain the variety of width and length of different burrows. Burrows for raising the young, nursery burrows and burrows for submergence at sunset, contain one or two nest chambers that are filled with dry grass. These chambers tend to be around 30 centimeters high and 50 centimeters high, these chambers are shown to be used for raising the juvenile and are hypothesized to be used as sleeping chambers for the adults. There also exist smaller chambers that are not used for these purposes, they can usually be found around 1 meter below the surface and are thought to function as turnaround points or temporary hideout that allows for listening to threats from above the surface. The sharing of burrows is very common, not only from generation to generation but members of a coterie also have access to a multitude of burrows.

There are a couple of predators that hunt prairie dogs, these predators can either hunt on them from above the surface or below the surface. Examples of above the surface predators are coyotes, bobcats and aerial predators. Snakes, American badges and black-footed ferrets hunt on prairie dogs from below the ground. Little is known about prairie dogs that die from natural conditions, but evidence suggests that this mostly happens below the ground and that dead bodies may be pushed out of the burrows by the other prairie dogs. There are also cases where bones where found in burrow mounds, they may have been carried above the ground when excavating or renovating burrows.

Burrow mounds have some interesting functionalities. First off, burrows help the prairie dogs protect against flooding rainstorms, which can occur in the areas they tend to live in. Prairie dogs have furthermore been seen to frequently run to the mounds in times of suspected danger, this is thought to be because the mounds allow for efficient scanning for predators. Most fascinatingly, burrow mounds stimulate Bernouilli's principle, thus creating better ventilation. This may especially be important in longer burrows.

Bronnen:

https://en.wikipedia.org/wiki/Prairie_dog (eerste deeltje) https://en.wikipedia.org/wiki/Black-tailed_prairie_dog https://books.google.nl/books?hl=nl&lr=&id=BJzzQXkka54C&oi=fnd&pg=PR13&dq=prairie+dog&ots=V21MgOvPXy&sig=bmR17lWn7fuCwlca53E9aXoREQY#v=onepage&q&f=false (black-tailed prairie dog)

Mapping

The robot needs to map out the burrows to improve their navigation. On top of that those maps could be analyzed for research purposes. Ferenc is looking into which sensor are required for mapping among other things and Mendel, Nick and Milan look into how the burrows can best be "visualized" for the robot navigation as well for analysis.

We first had a look at the slides of week 2 of the course Interaction with social robots (0LSUD0) where robot navigation is discussed. Mapping, among other topics, are brought up in the slides. According to the slides, there are three ways to represent an environment. These are using a continuous metric, a discrete metric and discrete topological. In a continuous metric, the world is described using coordinates (x and y) and angles (θ). In a discrete metric, the world is described along a grid. Exact angles are mostly lost, but relative positioning between locations and distances are mostly retained. In discrete topological only relative positioning between locations are retained.

The slides discuss two problems with map building, these are keeping track of changes in the environment and the representation and reduction of uncertainty of the robot. The latter is a general problem with robot navigation, but the first is of greater interest to us. Prairie dogs change their burrows over time, for example to accomodate more of them. This is a problem the robot is more likely to run into if it repeatedly has to explore the same burrow. On top of that, if the robot cannot tell the difference between the walls of the tunnels and the prairie dogs, then they will perceive them to be part of the burrow. Then when the prairie dogs move, which they will inevitably do, the burrow will seem to have changed from the robots perspective.

A System for Volumetric Robotic Mapping of Abandoned Mines describes two robotic systems designed for accurate volumetric maps of underground mines. The way these two designed move around is not of much interest to us as they are manually operated by humans. One of them even needs to be pushed forward by people, making the design useless for mapping out Prairie dog burrows. However, the technology used for the mapping itself is of interest to us. Both designs use laser range finders to sense the environment and create a 3-D map by firstly creating a 2-D map with a scan matching algorithm and secondly applying the same algorithm on 3-D measurements after the 2-D map has been completed.

These robotic systems are mainly designed for planar environments, but can still perform decently in non-planar environments by estimating the tilt of the sensors to reduce error in mapping. However, the authors do not discuss how much tilt the robotic systems can deal with. This is important for us to know as the tunnels dug by the prairie dog that our robot has to map out can be very steep. If there is a limit to how much the sensors can tilt before the error becomes to severe, then a better solution might be to modify, if possible, the algorithm such that it can deal with the extra dimension and angle. Depending on how exactly the robot will move through the burrows, modifying might be the prefered option as the robot could also have to deal with the possibility that it is slightly tilted along the other axis or even fully upside down. We have to better understand the actual algorithm to see if modification is even a possibility.

Probabilistic robotics. Communications of the ACM discusses the uncertainties that arises when determining the location of the robot and how to deal with them using probability distributions but does not go into detail how exactly the sensory data is processed and visualized.

brainstorming

mapping with the assumption that no animals are inside the burrow -there is a problem with this assumption: what if there are animals? --we can deal with that issue by installing some kind of detection system that is able to recognize whether what is detected is an animal or the wall of a tunnel in the burrow. ---one option for this is a heat sensor

---

what if an animal is diseased?

---

maybe attach a motion sensor

---

what if a prairy dog is diseased and the robot passes by, what will the other prairy dogs do?

Most underground robots use tracks to move forward. This requires the surface on which they move forward to be flat enough, but we cannot make this assumption, as the burrows of prairy dogs have tunnels that can be almost vertical. A robot with wheels that can be pushed to the side would deal with the changing width of the tunnels, but we need to be careful that such a robot might have difficulty turning. 2 Scissor like systems on either side of the main part of the robot with wheels on each end would reuslt in an 8 wheeled robot that would not fall over and would be able to find traction in all sizes of burrow. However, this came at the cost of maneuverability, it would be very hard to steer such a robot, and if the width of burrow tunnels varies too much then that means that the arms of the scissors would need to be very long or extendable. In Development of articulated robot for inspection of underground pipelines a robot similar to our first idea was discussed, it used 3 extendable arms with wheels on each arm to ensure the robot would be able to pressure fit itself to find traction. This was their first prototype which they had improved upon. The improved version had 2 modules. both consisted of arms that could extend with wheels at the end. The 2 modules are connected with a ball joint. This allows the robot to steer much better than the first design. This system was designed to explore pipes.

Fleas seem to be a problem for prairy dogs, we could look into counting those.

Sources

• Hyoukryeol Choi, Sungmoo Ryew, Sunghwi Cho, Development of Artiulated Robot for Inspection of Underground Pipelines
• Thrun, S. (2002). Robotic mapping: A survey. Exploring artificial intelligence in the new millennium, 1(1-35), 1.
• Thrun, S., Hahnel, D., Ferguson, D., Montemerlo, M., Triebel, R., Burgard, W., ... & Whittaker, W. (2003, September). A system for volumetric robotic mapping of abandoned mines. In 2003 IEEE International Conference on Robotics and Automation (Cat. No. 03CH37422) (Vol. 3, pp. 4270-4275). IEEE.
• Thrun, S. (2002). Probabilistic robotics. Communications of the ACM, 45(3), 52-57.
• Stachniss, C. (2009). Robotic mapping and exploration (Vol. 55). Springer.
• Karlsson, L. N., Pirjanian, P., Goncalves, L. F. D., & Di Bernardo, E. (2006). U.S. Patent No. 7,015,831. Washington, DC: U.S. Patent and Trademark Office.
• Thrun, S., Thayer, S., Whittaker, W., Baker, C., Burgard, W., Ferguson, D., ... & Reverte, C. (2004). Autonomous exploration and mapping of abandoned mines. IEEE Robotics & Automation Magazine, 11(4), 79-91.

Mechanics

To have a working robot, the mechanical parts and sensors are the most important and essential part of the robot. Without these the robot is rendered useless. The different parts also have to be chosen wisely as they need to be just right for the robot. For example the sensors need to be not only precise but also resistant to dirt and other small obstacles and the wheels need to be big enough for the terrain but small enough to maneuver.

To start off we will make a general overview of all the parts such a robot should have, and after deciding on a design, another list with the specific parts will be made. The robot should have wheels and a way to accelerate the wheels, which is possible with either multiple motors or with a central motor and a drive shaft. The robot should also have tires or spikes so that it has enough grip to move forward. The steering of the front shaft should also be possible in the robot, else we would have a problem at corners.

Effort Table

Week 1

Name	Total	Break-down
Nick Reniers	7 hours	Introduction lecture (2h), Meeting discussing subject (2h), Studying papers and editing wiki(3h)
Jankatiri Boon	6 hours	Introduction lecture (2h), Meeting discussing subject (2h), Studied papers and editing wiki (2h)
Milan Hutten	7.5 hours	Introduction lecture (2h), Meeting discussing subject (2h), Studied papers (3.5h)
Mendel van der Vleuten	8 hours	introduction lecture (2h), small scale test code (2h), meeting discussed subject (1,5h), studying papers (2,5h)
Ferenc Sterkens	6 hours	introduction lecture (2h), Meeting discussed subject (2h), reading some papers (1h), making planning (1h)

Week 2

Name	Total	Break-down
Nick Reniers	5 hours	Tutor meeting (20m), Group meeting (40m), recapping graph theory (2h), searching suitable algorithms and writing them (2h)
Jankatiri Boon	5.5 hours	Exploring and reading literature (3h), setup society and enterprise list (1h), write society and enterprise analysis (1.5h)
Milan Hutten	6.25 hours	Tutor meeting (20m), Group meeting (40m), Composing biological study list (30m), Looking for Interview Candidates (4.5h), Composing list of user needs and animal needs (15m)
Mendel van der Vleuten	5 hours	Tutor meeting (20m), Group meeting (40m), Exploritory literature research (2h), Studying papers of robots (2h)
Ferenc Sterkens	5 hours	tutor meeting (20), Group meeting (40m), Looking for interview candidates (3h), mailing candidates (1h)

Week 3

Name	Total	Break-down
Nick Reniers	10h	tutor meeting (20m), group meeting (10m), reading literature on prairie dogs in general (2h), decided to focus on black-tailed prairie dogs and read more in-depth literature on them (4h), filtering relevant information and compactly summarizing it (3.5h)
Jankatiri Boon	6h	Tutor meeting (20m), Group meeting (10m), Research users (3h), Page lay-out brainstorming and creating tasks for next week (2.5h).
Milan Hutten	9.5 hours	Tutor meeting (20m), Group meeting (10m), Research into mapping out the burrows (9h)
Mendel van der Vleuten	9.5 hours	Tutor meeting (20m), group meeting (10m), Research into prairy dogs (2.5h), research into mapping (4h), researching methods of tunnel traversion (2h), research relation burrowing owls prairie dogs (30m)
Ferenc Sterkens		tutor meeting (20m), group meeting (10m),

Week 4

Name	Total	Break-down
Nick Reniers		Tutor meeting (20m), Group meeting (2h),
Jankatiri Boon	2h20m	Tutor meeting (20m), Group meeting (2h),
Milan Hutten		Tutor meeting (20m), Group meeting (2h),
Mendel van der Vleuten
Ferenc Sterkens