RoboRat

Two years ago I published an article about creation of my robot Robutek II. I also promised some kind of “open-end” series, but unfortunately only the first article was written. The good news is that I had at least more time for playing, so there are many new things on the robot. There are actually some stories on robotika.cz revealing some of her history. For example in the robotics diary of the team Short Circuits Prague (Eurobot 2004 contest). The robot appeared under nickname “RoboRat”. Because we got used to this name, I will use it in this article too. This nickname RoboRat comes from its current job. It works as an assistant robot for experiments with teaching rats in the Institute of Physiology, Academy of Science. It works there for the second year already.

In one test robot behaves as a predator placed in circular arena with diameter 1m together with rat, whose task is to avoid robot in safe distance. RoboRat does not see the rat. It has very simple strategy: it moves straight until it hits wall with active bumper, backs up, turns in place for random angle and moves forward again. The task of RoboRat is not as much interesting as the conditions under which it has to do the job. RoboRat works in very dirty environment (rats do not take any respect to small robots). Also there is no technically experienced operator who could immediately fix potential problems, so the robot reliability has to be very high. After several small modifications from the original design RoboRat could be proud of two years trouble-free operation. According to the institute test data it traveled more than 7km with three service checks. For it was slightly modified to become a secondary robot. Her main goal was to act as a moving obstacle and to block the opponent robot. It worked very well thanks to the tough construction.

The main structure material for RobotRat are aluminum bands 16mm wide and 3mm thick. According to my experience they are easy to cut, drill or bend. For the construction of such a small robot they are tough enough. In the original draft RoboRat parts were linked with screws M3 with threads directly in aluminum. On several stressed places they had to be replaced with bolts and nuts (very useful were the ones with plastic fill, which do not get easily loose any more), because the threads in aluminum were not very lasting.

The motion subsystem went though the biggest changes. Originally there were two motors from Merkur (Czech kind of Meccano) – the same as in Robutek I. It was quite sufficient for the movement. Also the same tested electronics originally for Robutek I could be used (H-bridge L293D as a motor controller). Then I got the offer from the institute, and one of the requirements was different speeds of operation. The slowest required speed was not obtainable with these motors using PWM regulation. So I decided to use different motors.

The most affordable motor with gear, which is necessary for the slow speeds, is ordinary model servo. I cut position limits from the last toothed wheel and I fixed potentiometer so it does not move together with the wheel any more (the potentiometer is used by servo electronics to measure current wheel orientation). In the new setup, the servo will never reach desired orientation (potentiometer does not move) and so it continuously turns.

Another servo feature is that further positions (differences) are realized with higher speed. So if we fixed potentiometer in the approximate center position, we can control speed and direction with standard servo signal. So if we use 8bit resolution for the signal (values 0 to 255), servo will turn in one direction for values smaller than 128 and in the other direction for greater values. The absolute difference from 128 defines speed. In the ideal world would servo stop for value equal to 128 – the trick how to get close to this is to set output signal to value 128 and then turn potentiometer until servo does not move. In the real world this value changes depending on batery level and so it is better to stop servo simply with not sending any signal.

RoboRat is controlled by microchip Atmel AVR AT90S4433. The board has connector for serial communication UART, ISP programming connector, digital and analog I/Os and connectors for servos and encoders. The board also contains comparator LM393 with trims for digitalization of analog signals from encoders.

Encoders are the second part with the biggest changes in RoboRat evolution. Originally I used proved configuration where IR-LED with phototransistor (CNY70) scans robot wheel with black and white strips. This solution worked fine in normal circumstances. Unfortunately in very dirty environment it quickly lost original reflective property. Another problem was that experiments with rats are recorded by IR camera, which was disturbed by these IR sources. Moreover this setup was very sensitive to mounting and once a while somebody inexperienced changed mechanics with bad side-effects. This was motivation for different, more compact and robust, encoders. The final solution was placement of encoders directly into servo.

The goal was to place sensor CNY70 into servo and instead of striped wheel we could use one of toothed wheels. The whole encoder is then safely placed inside servo and is practically unbreakable, the surrounding does not influence its behavior and also emitted IR does not influence the environment. For placement I used space where usually is foam for fixing potentiometer (if needed, potentiometer axis could be also shorted). This space is sufficient for sensor CNY70 and few necessary resistors. I drilled a hole into the very first toothed wheel near the motor, so the sensor shines through the hole or not. For better reflectivity I pained bottom part of the wheel with white. I glued sensor and resistors for better compactness using hot glue. It is important that the hole is big enough so that both active parts are visible. I could use power (+5V) and ground from servo board thanks to fact that pull-up resistor and resistor for powering LED were both placed into the case of servo. So there is only one extra output cable from servo, which can be connected to analog input of microchip. Wiring of CNY70 is described in article Jak programátor k robotovi přišel.

The drawback of this solution is that it is not possible to detect direction of servo rotation. For this we would need two sensors but they would not fit into the servo case. In reality it is not a big issue because we typically know which direction should servo turn and spontaneous motion is exceptional. Servo motor turns very fast and has high torque. Note, that as a programmer it is necessary to pay attention at signal oscillations on the edge of the hole. It could easily cause “virtual” motion of robot. The resolution of servo is around 100 ticks per revolution due to fact that we use the fastest moving wheel.

Another requirement which arose during experiments was speed control (change speed within 20 minutes testing sessions). It needed wireless solution because cable would interfere with the experiment setup. The cheapest solution I found was one-way radio modules RTF-DATA-SAW and TX-SAW from Aurel. They are easy to buy (in CR you can find them in GES for around 400 Kč total i.e. 14EUR). Using relatively simple circuit from datasheets for these modules I could create boards. The transmitter board contains chip AT90S1200 which is connected with an old joystick. The receiver board is then connected directly to the control microchip on RoboRat. One timer of RoboRat board is used for signal sampling. Those two modules do not support any protocol for the data transmission – what is written to one pin on the first board will show up on another pin of the second board. Using relatively simple protocol I was able to transmit reliably data inside room within couple of meters (without any knowledge about antenna construction and communication media in general…).

Back view

Side view

Front view

Top view

Support

Mode selector

Inside

Bottom view

Original encoders

Servo wheel

Servo

After modification