Tweet
I was considering this for a UK ant so I'll put down my thoughts.
My first thought it that being an ant there was no way I could get sufficient processing and sensers on it to fight effectively. SO I started going down the offboard control route. I settled on a kinect 2 camera. Offboard gives a much more stable datastream than a robot being thrown around the arena and much easier to process stuff like the robots position in the arena. The 3D imaging capability of the Kinect 2 simplifies the image processing methods needed as you can reasonably assume a non flat bit of floor is probably a robot/hazard or wall. This gets around bots that might be very hard for the system to visualise. It also would have allowed me to throw a much bigger chunk of processing power at the problem using a desktop as opposed to a single board PC.
The eventual plan was to minimise the information needed for the robot to feed back to the point none is needed. Effectively this would allow the control system to control any bot not just one built for autonomy provided it had some kind of tactic for each.
The only issue is kinect 2 cameras can have trouble seeing through polycarbonate so there might have been the possibility of having to place the camera within the arena, not much risk in ants but more so in higher weight classes. On the plus side you can rig it to capture some nice battle footage too!
My first thought it that being an ant there was no way I could get sufficient processing and sensers on it to fight effectively. SO I started going down the offboard control route. I settled on a kinect 2 camera. Offboard gives a much more stable datastream than a robot being thrown around the arena and much easier to process stuff like the robots position in the arena. The 3D imaging capability of the Kinect 2 simplifies the image processing methods needed as you can reasonably assume a non flat bit of floor is probably a robot/hazard or wall. This gets around bots that might be very hard for the system to visualise. It also would have allowed me to throw a much bigger chunk of processing power at the problem using a desktop as opposed to a single board PC.
The eventual plan was to minimise the information needed for the robot to feed back to the point none is needed. Effectively this would allow the control system to control any bot not just one built for autonomy provided it had some kind of tactic for each.
The only issue is kinect 2 cameras can have trouble seeing through polycarbonate so there might have been the possibility of having to place the camera within the arena, not much risk in ants but more so in higher weight classes. On the plus side you can rig it to capture some nice battle footage too!
Comment