Welcome to The Neuromorphic Engineer
The Telluride Grand Challenge Robot Race

PDF version | Permalink

Timmer Horiuchi and Giacomo Indiveri

1 December 2005

Many and varied contraptions competed to be the fastest at escaping from the arena.

This year's Telluride Neuromorphic Engineering Workshop (June-July, 2005) featured a new event: the Telluride ‘Grand Challenge’ Robot Race. Our goal was to create a fun, focused activity that would guide the many different participant and workgroup projects towards a concrete demonstration.

Our chosen task was a combination of goal-seeking and obstacle avoidance behavior that challenged robots to negotiate an unknown obstacle course autonomously. We encouraged the integration of many different sensor modalities (e.g., visual, olfactory, auditory, tactile), decision-making models, and locomotion platforms (wheeled, slithering, and walking robots) to provide diversity and encourage innovative solutions. The competition took place in a simple rectangular arena with the start and finish points in opposite corners, and with obstacles (boxes and trashcans) placed at the last minute.

The rules of competition were straightforward:

Robot entries must run ‘autonomously’. Judges may allow a reset intervention for up to three tries.

‘Dumb’ beacons of various types may be used only at the end point. Any beacon you create must be considered public (i.e., usable by everyone, if desired).

At least 1/3 of the original mass of the robot entering the race must cross the finish line.

Robot entries may not damage the racecourse, judges, or bystanders.

Water, pyrotechnics, and radiation sources should not play any role in the robot's operation or construction (see also rule 4).

Obstacles will be finalized (but not their position) and shown to the competitors at least two days prior to the race.

All robots must begin with zero initial kinetic energy.

Robots must not move obstacles out of their initial locations.

In support of the competition, the companies K-Team and Wow-Wee Toys generously donated development and robot hardware, both for use in the competition and as prizes. K-Team sent the KoreBot development board, compatible for mounting on Koala mobile robots, but also good for general use. WowWee sent RoboSapiens and RoboRaptor platforms. Two RoboSapiens were used in the competition while we reserved a brand-new RoboRaptor as a prize.


Garcia, the Antennanator


Competitors

By racetime we had 12 entries that ranged from simple locomoting devices with little to no sensing capabilities to fully sensor-equipped robots with goal-sensors and obstacle avoidance behaviors. There were many notable entries:

Garcia, theAntennanator (Noah Cowan, Mitra Hartmann, Jennifer Laine, Vincent Chan, and Kate Williams) was a robot platform that employed two whisker sensors to detect obstacles and trigger movements that allow extended wall-following behavior. This robot succeeded in progressing about 80% of the way towards the finish line, negotiating two box obstacles before getting stuck against the wall. They did not implement any goal-sensing capabilities.


The Kave/Koala Team

Kave/Koala robot (Olivier Rochel, Peng Xu, Christian Faubel, André van Schaik) was a binaural cochlea chip that generated AER spikes from two microphones mounted on a Koala robot platform. For stimulation, they used a sonar ‘ping’ beacon at the finish line, and the spikes were used to generate an estimate of the interaural time difference of the sound. The Koala robot already implements infrared-based obstacle avoidance.

RoboSensation (Tobi Delbruck, Hisham Abdalla, Jonathan Tapson, Deborah Gunning, Gong Boonsobhak and Joseph Lin) was a RoboSapien platform that used a wireless network device to send commands and data between the robot and a laptop. The built-in microphones were used to estimate the beacon direction. Both a visual tracking chip and a commercially-available solvent sensor were interfaced to use an odor plume for goal guidance.


RoboSensation


Kave/Koala.

AudioSapiana (Antje Ihlefeld, Mounya Elhilali, Malcolm Slaney, Nima Mesgarani, Tara Hamilton, Jonathan Tapson, Stephen David, Sue Denham, David Anderson, Shihab Shamma, and Andreas Andreou) was constructed from a network of two computers that communicated sound data through the workshop's WiFi-accessible file system.1 (Full details in article on next page-Ed.) Using a binaural auditory saliency model, the two functions of goal-direction finding and recognition of the beacon sound were handled across two computers. This team suffered the ill-fortune of a town-wide power failure that put the computer network in a funny state. After getting the system functional again, the robot headed out, but did not operate properly. (There is video of the robot elegantly completing the course soon after the competition was over.2


Noah J. Cowan.

RoboCup (Hisham Abdalla) was a hand-built robot that detected ultrasonic signals transmitted by two beacons in the room to estimate x-y position in the room. This allowed the system to virtually ‘sense’ the direction of the finish line. A prominent feature of this robot was the plastic cup used to hold up one of its ultrasonic transducers.

It (Chris Twigg and David Graham) was a small toy car on which two binary whiskers were installed. This car achieved a high speed and would turn when one of the whiskers ran into an obstacle. When both whiskers are deflected, a reverse command was issued after some delay. The car managed to careen off of two obstacles and strike a wall extremely close to the finish line.


And the winners were…

The two main prizes, provided by K-Team and WowWee, were won by Garcia, the Antennanator, for the ‘Coolest Approach Attempted’ award, and by Kave/Koala, for the ‘Most Successful Robot’ award. Additional awards were given for the following categories: ‘Fastest Entry’ was It; ‘The Rube Goldberg Contraption’ was RoboCup; ‘Best Looking Robot’ was AudioSapianna; ‘Most Pathetic Performance on the Field of Battle’ was RoboSensation; and the ‘Most Integrated Robot’ was RoboSapianna.


Closing thoughts

Overall, we feel that the robot race was a great success in drawing the interest and efforts of many of the participants. This year we saw the emergence of the RoboSapien as a new robot platform and the effective use of wireless network devices in various systems. We also received lots of feedback that perhaps too much of everyone's time was spent in making the most basic hardware operational, leaving less time for algorithm development and systems-integration issues. In future competitions, we intend to provide more infrastructure and ready-to-use robotic and development platforms to encourage the use of common locomotion platforms and communications. We encourage all of you to contact us about your thoughts on the competition and ways to make this an even better event for the future.




Authors

Timmer Horiuchi
Department of Electrical Engineering Institute for Systems Research Neurosci. and Cognitive Sci. Program, University of Maryland

Giacomo Indiveri
Institute of Neuroinformatics, Uni/ETH Zurich


References
  1. A. Ihlefeld and M. Slaney, The story of AudioSapiana, The Neuromorphic Engineer, 2005.

  2. http://www.theredplanet.co.uk/Telluride2005


 
DOI:  10.2417/1200512.0028

@NeuroEng



Tell us what to cover!

If you'd like to write an article or know of someone else who is doing relevant and interesting stuff, let us know. E-mail the editor and suggest the subject for the article and, if you're suggesting someone else's work, tell us their name, affiliation, and e-mail.