DGC Follow-On Discussion Notes 2005-11-23

Summary of the discussion of 11/23/05
Pietro Perona, Richard Murray, Stefano Soatto

Rationale The DARPA Grand Challenge was designed to push the envelope of autonomous vehicles for defense applications and gathered a significant amount of attention from competitors and teams. From many perspectives it has been a success. Part of the success may stem from the novelty, the prize, and the significant organizational effort by DARPA. On the other hand, the design of the rules (staggered departures, no moving obstacles, off-road environment) and the timing made it such most of the effort went on building the mechanical and electrical systems (in some cases the entire vehicle was built from scratch) rather than pushing the design of sensing and algorithms. Association with the US military made international participation, and direct involvement of automotive manufacturers, problematic, although there were notable exceptions.

Resolve To design a competition that fosters advancement in autonomous driving for civilian transportation, including urban environments with pedestrian, cyclists, including varied environmental and weather conditions (day, night, rain, snow etc.). The goal is to ease the societal burden of civilian transportation in terms of both safety (1.5M deaths in traffic accidents worldwide, 40K in the US alone) and productivity (releave congestion, shorten commuting time, allow drivers to perform other tasks while driving).

Given the interest generated by the DARPA Grand Challenge, given the difficuty in conducting such a competition in populated areas, and given that many team have ready-to-race vehicles with nowhere to go, one option is to design a series of races that gradually move from DARPA Grand Challenge scenarios to fully populated urban scenarios in stages. Richard Murray indicates that such a plan was put in place by the National Federation of the Blind Jergen Institute before the second DARPA Grand Challenge was issued. A possible sequence of competitions could be:

1. Desert, blocked to traffic, similar conditions to DARPA Grand Challenge, but with all vehicles taking off at the same time, few GPS waypoints. No restrictions on cost or uniformity of equipment. Traffic laws and responsibility rules woudl apply, with a "safety margin" (i.e. a vehicle coming within a 2m radius from another contestant would be consider a collision and trigger e-stop by the offending vehicle). The race would be conducted on a smaller course, most of it visible from a grandstand. Access to the event would have to be regulated (through a qualifying event similar to DGC?) which would run significant cost.
2. Fake urban scene (in one of the many ghost-towns of the southwest), maps available, A-to-B type of goal, no intermediate GPS waypoints. No restriction on cost or equipment. This would decrease the emphasis in off-road environment. Organzational challenge, in addition to qualification, is to find and prep suitable venue.
3. Semi-urban scene: same as 2) but with moving obstacle (tele-operated pedestrian phantoms). Same challenge as before.
4. Urban environment: same as 3) but on a closed-off section of a few city blocks. Challenge would be to find suitable venue. It is not rare to have sections of downtown LA closed off to traffic during weekends to shoot movies, but closing off a large enough section may have significant cost.

Additional comments:

• The competition could have "cost categories" so that a winner is declared in each category, say the 10K\$ category, the 50K\$, 100K\$ etc. to promote and enable commercially viable solutions for autonomous driving.

• Direct involvement of a car manufacturer staring in 3) would allow all competitors to compete with the same mechanical/electrical platform (as in a "Toyota Prius Cup") allowing teams to focus on development of new algorithms.

• Pietro Perona suggests a formula like the America's cup, where the winner decides the timing, place and format, and the challengers adhere in a gentlemanly manner. In this case the organizational burden is on the winner, but it comes with the benefit of designing the competition and the ensuing visibility. In this format, the competition would likely travel outside the US, with resulting increased operational costs and logistics.