Embedded Motion Control 2016 Group 7

Group Members

ID-Number	Name	E-mail
0816253	Emile van Halsema	e dot v dot halsema at student dot tue dot nl
0805999	Nard Strijbosch	n dot w dot a dot strijbosch at student dot tue dot nl
0816608	Raymond Kerstens	r dot j dot c dot m dot kerstens at student dot tue dot nl
0814199	Frank Heck	f dot j dot m dot heck at student dot tue dot nl
0816361	Ids van den Meijdenberg	j dot w dot a dot v dot d dot meijdenberg at student do tue dot nl

Journal

In the journal the progess of the group will be described

Week 1

In the first week challenges ware specified and concretized. In addition every member installed ubuntu and read up on c++.

Week 2

In the second week we dove into code, a model was written for determining the nodes for corners and intersections, for preventing PICO to hit the wall and for plotting a course trough the corridor. In the simulator this seemed to work, however at the experiments we experienced multiple problems such as 'Ghost nodes' in large gaps, small gaps in the corridor and different values for the minimal angle, maximum angle and number of nodes in the laser data. Also when PICO should be going straight, it is drifting away. During the tutor meeting in week 2, we ware advised to start looking from the top, instead of from the bottom and start looking to the challenges as optimization problems. The lecture on wednesday also reflected this.

Week 3

From the feedback from the tutor meeting we concluded that we got ahead of ourselves and should take a step back to look at the problem from above. With this approach we created the presentation which was held at the end of week 3. For the corridor challenge we devided the model in the sections as thought of in the composition patern. To avoid problems encountered in our previous experiments first préprocessing is done. A function is written that eliminates the 'ghost points', all points beyond a trust region are set to a constant value, and the number of points saved for processing is reduced based on distance between the points. A new optimization function is created which searches for the open space in the obtained laser data using rectangles. The resulting rectangles can then be translated to the possible options in the maze(for example, a left corner, right corner, intersection or T-crossing). For the corridor challenge the only possibility is the T-crossing, in which case it should take the first exit. The resulting rectangle information is then used in the next function which is the setpoint. The setpoint function calculates 4 points in order to take the corner at the crossing. Whenever PICO is near a setpoint, the next point is taken such that it will go trough all points. In addition to the setpoint a potential field pushes PICO away from the wall when it gets to close.

Initial Design Embedded Motion Control Group 7

The project is dived in two parts: the corridor challenge and the maze challenge.

Corridor challange

For the corridor challenge the goal is to take the first exit.

Requirements

For this corridor challenge a number of requirements have to be met for an arbitrary corridor:

• Reach exit as fast as possible.
• Be able to recognize walls.
• Don't hit the walls.
• Be able to recognize and take the first intersection.
• Be able to rotate and translate.
• Be able to recognize maze exit.
• Stop when exit is reached.

Specifications

These requirements can be rewritten to specifications:

• Reach end within 1 minute.
• Recognition walls and intersection.
  • Objects in straight line (with variation under 5 cm) are considered walls.
  • Opening of at least 0.5 m is an intersection.
• Don not hit walls.
  • Distance to walls at least 10 cm.
• Be able to drive.
  • Drive in direction of point (updated 30 times a second).
  • Maximal 0.5 m/s during straight line.
  • Maximal 1.2 rad/s rotation.
• Recognize exit and stop
  • Drive straight for 3 seconds after being 30 cm from intersection, then stop.

Functions

In order to meet these specifications the system is decomposed into multiple functions:

• Intersection recognition: identifying where the intersection is.
• Path planning: calculating the optimal path through the corridor.
• Safety features: Ensuring that PICO does not hit the walls.

Maze challange

For the maze challenge the goal is to find the exit in a maze. The location of the exit is unknown, there might be loops and there is a door in the maze.

Requirements

The requirements of the corridor challenge also hold for the maze challenge but are elaborated with:

• Be able to map the maze.
• Be able to locate itself in map.
• Be able to recognize maze components (doors and walls).
• Be able to open a door.
• Be able to detect the maze exit.

Specification

These additional requirements can be written to specifications:

• Mapping
  • Update map 30 times per second.
  • Remember path.
  • On the driven path map walls, doors, crossings.
• Be able to localize itself in map.
  • Localize position with precision of 5 cm.
• Recognition possible doors.
  • Recognize dead end of at least 30 cm.
• Open door.
  • Beep when in front of detected door.
• Recognize maze exit.
  • Stop when no walls are detected in front.

Functions

The functions from the corridor challenge can then be completed to obtain the functions required for the maze challenge:

• Mapping: Save all intersections, doors, and connections.
• Optimal control: determine the best strategy for completing the maze

Interfaces

o keep track of the states in which the robot is, the states will be printed in the screen. Also is the idead to show the mapping for the maze challenge on the screen for debugging.