European Urban Challenge Contestants Claim Their Competition Requires “More Autonomy”

You’ve no doubt heard about the DARPA Urban Challenge in which autonomous vehicles navigate through a simulated urban environment to meet various objectives. A lesser-known (to Americans) competition is being held in Germany with similar ambitions; however, those guys are talking smack about our American robots by claiming that their competition “demands more autonomy.” As if.

The European Land Robot Trial is Germany’s version of the DARPA Urban Challenge. In it, robots are expected to navigate 8km (a piddly 4.9 miles compared to over 100 miles for the Grand Challenge). An article in the EETimes claims that “the European Land Robot Trial does not allow the intensive use of satellite navigation.” That’s incorrect. According to the rules taken from the Military European Land Robot site:

3.9 Position Determination Signals

Challenge vehicles may be equipped to receive and process electronic position-determination signals (such as GPS) that are openly or commercially available to all teams. Position-determination signals that are neither openly available nor commercially available to all teams are prohibited.

Any costs associated with any subscription service are borne by the team.

GPS signals might not be available throughout the route at all times. GPS alone will not provide adequate navigation information to a vehicle. There will also be dust, smoke, and other visual obscurants on the route, and visual-spectrum-only sensing may not be adequate under these conditions.

This language is very similar to the DARPA Challenges’. The idea is that on occasion GPS may not be available due to valleys or tunnels. At those times, it’s very typical to use a combination of Kalman algorithms and inertial sensors combined with basic odometry to estimate dead reckoning positions until GPS is available again. Every robot at the 2005 DARPA Grand Challenge had to demonstrate this ability.

There is little in comparison of difficulty between ELROB and the Urban Challenge. The rules also state:

3.1 Autonomous Vehicle Behaviour

Participating vehicles might require autonomous behaviour and operation to complete ELROB. Fully remote controlled UGVs/UAVs are also allowed. Nevertheless, Vehicles must be unmanned.

Are they kidding me?! Remote controlled “robots??!!” “Might” require autonomous behaviour? I didn’t see this in Mr. Hammerschmidt’s article. In the event somebody actually does develop an autonomous robot at ELROB, there is a clarification in the FAQ‘s regarding the use of GPS:

Will it be possible to get more information on the route (e.g. GPS data) before the event?

No.
Also it is not possible and not allowed to get on the trial site before the event.
Any violation (even the attempted) will be published and result in exclusion from the event.
However on the first day there is a inspection of the tracks (all participants; by foot only, no robots or vehicles) where you can use cameras, GPS, etc. to gather data.

If during the Grand Challenge of ’05 we had been allowed to walk the course (and it was less than 5 miles) we could have dropped GPS waypoints in interesting areas to identify obstacles or best paths.

The fact is that there is absolutely a right place for image-based navigation. In the event there are no features in the path (such as a wide flat uniform expanse of packed sand) one might rely on image-based localization by scanning for recognizable patterns on the horizon. But frankly, in the event GPS satellites are dead, we’ll be facing bigger problems than a broken robot.