Another one of the house robots that wanders around is TRAX. TRAX is based on the original version 1 Traxster from
Robotics Connection in the US.
The unit is a two motor tracked base, modified to put sealed bearing races on the rear axle (the original steel shaft through aluminium was a bit squeaky). The drive units are the original gearhead units that came with the unit complete with in-built encoders, and the whole set-up is driven with a Serializer board.
The Serializer controllers are quite unique in our robot collection as they off load the processing to a PC via a wired or wireless link. The program on the PC is written in .NET and uses a simple text based command set, all encapsulated in the .NET library. The Serializer in this unit is a version 2 board and has a bluetooth interface daughter board for the connection back to the processing unit. This gives flexibility on the controlling host as it can be any device that supports a RFCOMM connection over bluetooth.
In addition to the base mobile platform TRAX also has a number of sensors and subsystems. The primary collision avoidance system is based upon a SRF08 sonar sensor at the front and a Sharp IR distance sensor at the back. The Sharp sensor is interfaced via one of the analogue ports on the Serializer, but all other modules make use of the I2C bus on the processor board. This has allowed for the inclusion of a magnetic compass (
Devantech CMP S03) and a text to speak unit (
Devantech SP03).
The unit also has a pan and tilt mechanism that carries a CMUCAM2 camera. This camera is an all-in-one vision system utilising a serial control protocol. As the Serializer only has one serial port, the CMUCAM is linked to the host PC via its own bluetooth connection, using an old BlueSmirf bluetooth serial dongle from
SparkFun. The only operation we have had any issue with is grabbing the complete image over the link. The data sent seems to reset the bluetooth module. There are two solutions to this issue:
- Don't grab an image as it isn't actually needed for the processing as that is all done on the CMUCAM.
- Set the unit to fast mode where all data is treated as data and no command sequences are looked for.
Currently, the CMUCAM has its own .NET library, developed in C#. So far I have implemented all of the configuration commands and also the colour tracking capability. Future plans include frame differencing for motion detection and histographic functions for increased edge detection capability.
The primary goal of the TRAX platform is to provide a system for investigation of basic computer vision systems, in this case via the CMUCAM.