ROBOTS IN HUMAN ENVIRONMENTS (RHUM)


2015-2016 Research Project Proposal

Title: Robots Among People, a Bio-Inspired Visual Navigation Approach

Description:

The purpose of crowd simulation is to simulate crowds of individually walking humans. Crowd simulation which is related to Computer Animation has been an active research domain in the past twenty years. Numerous walking models have been proposed over the years. In all cases, the primary purpose is to obtain visually realistic behavior for the crowd as a whole. To some extent, the fact that collisions between walkers could take place occasionally is not a big problem since everything happen in a virtual world. In a real world things are different and if an actual walker, say a robot walker, were to be controlled using the techniques proposed in crowd simulation, it would be highly desirable that said robot walker would not collide with the humans around. The purpose of this project is to investigate this issue and explore to what extent crowd simulation techniques can be utilized to control a robot designed to move among humans.

Fig. 1. The Robulab mobile platform. Fig.2. The Kinect vision and range sensor.

The target crowd simulation technique is the one that have been proposed in [1]: it is a technique that relies on the same visual cues that a human uses when moving round among static and moving obstacles. These visual cues are (i) the time derivative of the bearing angle and (ii) the rate of growth of a given obstacle. The goal of the project is to implement this crowd simulation technique on a robot platform such as the one depicted in Fig. 1 and to assess whether it allows safe navigation in real dynamic environments. The visual cues will be extracted from the optical flow and disparity map computed on the output of a vision and range sensor such as the one depicted in Fig. 2. They will then be used to steer the mobile robot. Besides the literature review, the following two functions have to be developed:

  1. Visual cues computation.
  2. Visual cue-based navigation.

They will have to be integrated and evaluated in simulation and on the real robotic platform.

References:

  1. Jan Ondrej, Julien Pettré, Anne-Hélène Olivier and Stéphane Donikian A Synthetic-Vision-Based Steering Approach for Crowd Simulation. SIGGRAPH 2010.

Contacts:

Requirements:

Interested candidates (Master, Engineers or PhD Students) pursuing a degree in Robotics, Electrical Engineering, Artificial Intelligence, Computer Vision, Computer Science, or other related fields are invited to apply for admission. Fluency in English is required and knowledge of French is a plus. The candidate should be experienced in object-oriented software development (C++, Java). Any additional experience in Robotics or Computer Vision is welcome.


Valid XHTML 1.0 Strict Valid CSS!