Programming the robotic dog AIBO Semester Project

1
Programming the robotic dog AIBO
Semester Project

• Michaël Gerber
• Assistant Sarah Dégallier
• Professor Auke Jan Ijspeert
• Biologically Inspired Robotics Group (BIRG)

2
Goal of the project

• Analyze several algorithms used for path planning
• Implement choosen algorithm
• Make Simulation with several robots using Webots
  in a dynamical approach
• Experiment with the real Aibo

3
Probalistic roadmap algorithm

• Randomly generate points
• Connect the points if they are reachable
• If there are several connected components, start
  first step
• Connect the start and final point to the roadmap

4
A algorithm

• Start with green node as current node
• Look at all neigbours, store them in an open list
• If the neigbour is an obstacle, dont store it
• Look for the neigbour in the open list closer to
  the red node
• Put this neigbour in closed list to draw the
  trajectory
• Start at the first step but with the neigbour as
  current node

5
HPA algorithm

• Divide map into clusters
• Put nodes at the limits of the cluster
• Link nodes with thoses neighbour nodes
• Link the nodes within the cluster

6
Artificial potential fields algorithm

• Put positive potential field for an obstacle and
  negative potential field for the goal position
• The positive field repulses the robot and the
  negative field attracts it
• Compute for each point of the map the influence
  of the potential fields
• Find around the current point the one with the
  smallest influence
• Go to the latter and start again the previous step

7
Choice of the algorithm

• We have choosen the Artificial potential fields
  algorithm
• It uses very simple mechanisms
• It is easy to implement

8
Calculating potential

• Uneg(p) ½Cd(p)2 if d(p) e
• Uneg(p) eCd(p) ½Ce2 if d(p)gt e
• Upos(p) ½D(1/d(p) 1/?)2 if d(p) ?
• Upos(p) 0 if d(p) gt ?
• Utot Uneg(p) Upos(p)
• We compute all the points gk of the map with
  those equations

9
Smoothing path with dynamical equation

• xi the current position
• gk the point to reach
• a, ß constants
• Equation dx/dt a(x gk)
• Condition if x gk lt ß ? gk gk1
• We use those equations to compute the path
  between the points gk
• The smoothing path will be in blue in next slides

10
Matlab results
11
Matlab results
12
Future work

• Implement the method with Webots
• Simulation with AIBO using Webots
• Simulation with other robots (e.g. iCub)
• Experiment with the real AIBO

13
Simulation

• Implement the Artificial potential fields
  algorithm using Webots
• Implement the trajectory method to follow the
  path
• To turn, aibo must make a diagonal path with his
  foot

14
Question