Today, Thursday, we are scheduled for Round II of QID. Our analysis from data of Round I indicated that our bot may have been stopped because of two reasons (1) unsynchronized sensors and (2) too much disk writes. Patrick Landry, the system administrator, on the team ruled out reason number 2. He concluded that it was not the amount of disk writes, but a lose connector on disks that was the cause of the problem. Cause 1 is still critical though. If our sensor data are not synchronized, then we cannot relate data from multiple sensors. To give an example in human terms, if what the eyes see cannot be time-synched to what the hands feel, then we’d never be able to drink coffee. We have known the issue. It was not a priority before because we really did not have all the pieces together. The issue has become top priority now. Yesterday evening we decided to bite the bullet and rebuild the Linux kernel to introduce the necessary capability. The decision has its own risks. It is not easy to integrate all the hardware drivers (for joystick and other sensors). As one would expect, we are now running into the problems. As I write this, Patrick is sitting outside the garage on the asphalt, trying to find the reason why the laser range finder (from SICK) is not working. We have spent enormous amount of time and money on these laser range finders. Every time we do any changes, they seem to break. But it appears we have no choice. All other teams are using the same product. Coming back to the building kernel, Team CajunBot has been running against time from Day 1. So it is no surprise that we will take the risk a few hours before QID Round II. Besides, we need the new capability to improve our system. And there is no real point in doing a test run with a system without adding new capabilities that improve upon previous run.
Arun Lakhotia