Crisis -- the plot thickens!

Well,  at this point (into May!) a number of problems had matured into crises and we had some serious issues to overcome.

Early in the winter semester (when the students begin working on the project) we were worried about the software development environment and its support for algorithm creation and testing.  In the past we had done 95% of our coding in Matlab using a PlayerMex interface that we had developed to execute Matlab code within the Player environment.  We used C-MEX files for speed sensitive routines and regular matlab scripts otherwise.  This had proved to be a very effective code development environment due to the interactive nature of Matlab and the excellent debugging and visualization tools available.  

When we decided to switch to ROS we were committing to developing our code in C/C++.  This proved to be a slow process and our progress was inadequate.  Therefore, we begin to explore the IPC-bridge software available through the Univ. Penn. Grasp lab.  However, only a few ROS-message types were supported and we set out to create the necessary messages for image processing and mapping.  This work went slowly with many problems cropping up.  Ultimately we came upon a ROS-Java solution that was very compatible with Matlab's Java user interface.  This last direction permitted us to develop the image processing code in matlab but still capture and publish images from ROS.  We still wrote the mapping code, local navigation code, EKF code and global path planning code (D*) in C/C++.

While the software drama was unfolding, a more serious crisis had developed with the Husky hardware.  We were having difficulty with our navigation/driving programs and it became apparent that the Husky vehicle was behaving in a lethargic fashion and was exhibiting motor controller errors which essentially froze the vehicle.  It took a while to trace these problems down -- our motor temperatures appeared to be fine (nothing exceeding 52 degrees C when our limits were at 60C for safety and 70C for shutdown) but we determined that we were showing higher than expected currents.  

Ultimately we found (with help from Clearpath Robotics) that the motor armatures were overheating despite the fact that the case temperatures were not outside the limits.  The problem was that the drive train modifications implemented by Clearpath to meet our 2.2 m/s design criteria caused the average currents in the motors to be too high for the thermal transfer characteristics of the armature.  So we had a cascade failure with the windings which weakened the power output which futher increased heat. 

Burnt armature windings are visible -- note the vaporized insulation deposits!

Motor armature, brushes and field magnet casing

The Husky was Dead!  We worked feverishly with Clearpath to determine possible solutions, but Clearpath (to their credit) indicated that their chassis drivetrain modifications were in error and could not support our higher-speed driving specifications with the terrain and payloads assoicated with the IGVC.  The original Husky would function fine, but at a reduced speed of 1 meter/sec.  

Thus with one week until the competition we decided to attempt to transfer all the computer hardware, sensors, specialized power systems, and code to one of our former IGVC Chassis.

Clearpath Robotics was very helpful, providing for a return of the vehicle, and software support to develop ROS drivers for our "new" drivetrain.  They were planning on attending the IGVC, so we arranged to meet them there so that they could collect the damaged robot -- they will return it to the factory spec, and we will use it for our other classroom robotics projects (we have a bachelors program in Robotics and Mechatronic Systems and thus many associated course-related robotics projects).

Clearpath brought their new heavy duty Grizzley robot to the competition for demonstrations, and ended up using it to pull our stripped down Husky platform away!

MJP 0251

This browser cannot play the embedded video file.

Grizzly pulling our Husky away!

Here are some other pics of Revenant's (1) ignominious departure...


©Electrical, Computer, Robotics, and Mechatronic Systems Engineering -- UDMercy