# Robots minimum distance

I am trying to implement a mechanism to make robots avoid being too close (Say in a distance less than d). I am not familiar with those systems and I have to implement a strategy to avoid robots being too close to each other. Could anyone recommend me some readings for such a problem or a set of keywords to search for? I don't know yet how to start.

• This question needs more information to be useful. Are you talking about robot arms or vehicles? Does this involve using a sensor for detecting the other robot, or is the other robot's position provided to it? – Ian Mar 27 '13 at 18:18

Do you mean too close to each other?

It sounds like you might be referring to the sorts of steering behaviours that Craig Reynold's refers to in his Boids project. He describes some behaviours that exhibit as flocking or herding because of the way the participants avoid each other, yet stick together. He uses the terms 'separation', 'alignment' and 'cohesion' to describe the algorithms involved.

Craig applies these to computer simulations, but you could just as well apply the concepts to robots.

Other searches that may prove beneficial: 'flock simulation', 'crowd modelling'.

• I am also doing a simulation to be honest! I don't have real robots! I just didn't know what to search about the motion in order the robots to avoid each other ! – nikosdi Mar 27 '13 at 23:22
• thank you very much ! If i was able I would upvote this answer. It is really really helpful. – nikosdi Mar 27 '13 at 23:27
• To be clear, Boids is a theory about flocking behavior. It requires agents not to collide, referred to as separation, but it doesn't prescribe a method of control to keep them from doing so. – DaemonMaker Mar 28 '13 at 18:58

A good place to start is with the work by Dr. Jur van den Berg and his colleagues. Check out the publications velocity obstacles and reciprocal collision avoidance. You could start with the latest paper, Reciprocal Collision Avoidance for Robots with Linear Dynamics using LQR-Obstacles, they have released on the subject and use the citations to find more related material.

A very effective and elegant approach is the resort to the so-called Potential Fields.

Essentially, each robot of the swarm generates a repulsive potential field which is tuned to be limited in the space around it (pretty much as a charged particle), so as any specific target is responsible for an attractive field, which in turn can reach farther distances.

Then, the single moving robot accumulates the field in the space location where it is by summing all the contributions in sign. The resulting potential will be the driving force to be integrated for yielding the final motion.

To make things work easily, the negative fields must be set very strong but limited in the space around robot mates. However, the designer can have fun choosing the potential shapes and playing with tuning parameters.

I'd like to suggest that you start with a book or a course in principles of robotic motion for this topic.

Principles of Robotic Motion Provides a very detailed introduction to this topic starting chapter 3 onwards.

Alternatively, you might want to look for how to make efficient configuration spaces for robots to traverse.