2004 Research Project with CalTech Multi-Vehicle Wireless Testbed

UCLA Students:

Chung Hsieh - Department of Electrical Engineering, UCLA
Bao Nguyen - Department of Electrical Engineering, UCLA
David Tung - Department of Computer Science, UCLA

Mentors:

Robotics: Zhipu Jin and Ling Shi (CalTech graduate students), Richard Murray (Director of the Lab)
Algorithms: Dan Marthaler (boundary tracking), Yao-Li Chuang (virtual potentials), and Andrea Bertozzi (overall project)

Project 1: Boundary tracking

Project leader: Chung Hsieh

Each vehicle moves in a circular path that is either clockwise or counter clockwise depending on whether the vehicle measures itself to be inside or outside of the sensing area. We use a virtual sensor determined by feedback from the lab positioning system. The composite motion results in the vehicle traversing the boundary and collecting a sample of boundary point data. For multi-agent implementation, the vehicles slow down or speed up depending on whether another vehicle is close by or far away, in order to maintain some spacing between vehicles.

Link to more details-under construction! (see ACC paper below)

Project 2: Greedy search algorithm

Project leader David Tung

All vehicles have a list of targets to visit. Each vehicle has the same list. The vehicle chooses an initial target from the list, based on distance, and moves to visit that target. If another vehicle has also chosen the same target, the two vehicles discuss who will get there first and the farther one chooses another target on the list. This is called the ``greedy'' search algorithm. David Tung implemented this on the testbed using the Steelebots. As first order control vehicles, their motion can be fairly precisely specified and thus they serve as a good choice for this method. David wrote a point-to-point controller for the Steelebots that allows a given vehicle to start at point point and travel to the next point, making sure to break and slow down so that it does not overshoot the target point. In order to implement the algorithm on the Kelly, a second order dynamic vehicle, the point to point controller from Project 3 (below) was used. The Kelly implementation also made use of their peer-to-peer wireless communication capabilities. Their trajectories were computed onboard and information was shared directly from vehicle to vehicle for a more decentralized implementation of the algorithm.

Link to more details (also see ACC paper below)

Project 3: Virtual potentials

Project leader Bao Nguyen

This control algorithm is specially designed for second order vehicles for which accelerations can be well controlled but positions less so. The algorithm has two components of the motion. The first is self-propulsion in the direction of motion, the second is attraction and repulsion to other targets or vehicles using pairwise potentials. With the help of Yao-Li Chuang, Bao coded up the controller for the fans on the Kelly vehicle to mimic the accelerations resulting from virtual potentials placed at specified points. They use this method as a point-to-point controller for the greedy algorithm (above) for second order vehicles.

Link to more details-under construction! (see ACC paper below)


Papers submitted and project reports:

  • Experimental validation of an algorithm for cooperative boundary tracking, by C. Hsieh, Z. Jin, D. Marthaler, B. Nguyen, D. Tung, A. Bertozzi, and R. Murray, August, 2004, submitted to the American Control Conference 2005.
  • Experimental implementation of an Algorithm for Cooperative Searching of Target Sites, by D. Tung, B. Q. Nguyen, C. H. Hsieh, Y. L. Chuang, Z. Jin, L. Shi, D. Marthaler, A. L. Bertozzi, and R. M. Murray, 2004, submitted to the American Control Conference 2005.
  • Virtual attractive-repulsive potentials for cooperative control of second order dynamic vehicles on the Caltech MVWT, by B. Q. Nguyen, Y-L Chuang, D. Tung, C. Hsieh, Z. Jin, L. Shi, D. Marthaler, A. L. Bertozzi, R. M. Murray, submitted to the American Control Conference 2005.
  • Joint UCLA-Caltech Report on the Implementation of Multi-Agent Cooperative Control Algorithms , by Chung H. Hsieh, Bao Q. Nguyen, David J. Tung, Ling Shi, Zhipu Jin, Yal-Li Chuang, Daniel Marthaler, Andrea L. Bertozzi and Richard M. Murray, September 2004.
  • Caltech MVWT API by Jon Gibbs and David Tung, August 2004.