Not a member yet? Register for full benefits!

Username
Password
VR Interfaces: ACE (Autonomous City Explorer)

Overview

Overview of ACE (Autonomous City Explorer)
ACE is a product of the Institute of Automatic Control Engineering at the Technical University of Munich. It is basically an attempt, built up over the past 18 months, at assembling a self-learning robotic system that does not use simultaneous location and mapping (SLAM) to find directions, instead, asking those it meets and inferring their responses.

Such a system needs considerable skill in natural language processing, machine vision, the ability to create its own maps on the fly (very much like SLAM), but at the same time, be able to integrate into those maps, the knowledge of others, all around it.

On its inaugral mission, a week before this article was written, ACE was sent from the University gate, to to the Marienplatz in the centre of Munich, about 1.5 km away. To hinder progress, ACE was given no maps, no GPS, no nothing, just pushed out the door and told to find its own way there.

The robot made it, and only took five hours. During that time, it had logs of interaction with 38 different people, whom it had waylaid, and asked for directions. The robot was capable of understanding to some degree, the instructions it was given, and capable of a new paradigm ? posture recognition ? to judge from the stance of the person, which way to go.

The researchers, led by Martin Buss at TUM, said that ACE was making good progress until it came to a busy pedestrian area and 21 curious onlookers stopped to investigate and talk to the robot. This bogged it down considerably, as it tried to ask directions, but the responses of so many people, more interested in looking at the robot, than helping it, confused ACE greatly. It did however manage to correct its errors, with the next individual person it waylaid.

The project goals of the ACE system, are to hopefully produce a robot which operates efficiently and safely in human urban, populated environments. At the same time, able to co-operate and interface with humans at both a high, and natural level. Obviously, there is some way to go yet, but every trip ACE makes, brings us closer to that goal.


Untitled Document