Teleworkbench: A Remotely-Accessible Robotic Laboratory for Education

We have designed a teleoperated platform for managing multirobotic experiment. We call this system Teleworkbench [6]. The aim of this system is to provide a tool for managing and analyzing experiments involving one or many minirobots. Thus, by using this system we aim for more e cient resource utilization, in this case robots, and at the same time for providing more access to robots regardless of users location. There are some unique features of this system: the ability to download a user-de ned program for each robot, the support for tracking more than 30 robots simultaneously, live video of the experiment, and the visualization of occurring events in experiments. To date, there are several similar attempts to broaden access of mobile robots for education purpose. One example is the work of Messom and Craig [3]. They designed a web based workbench for doing experiments with PID controller of real mobile robots. To prevent heavy bandwidth, this system sends only the position information of the robots instead of image or video data. On the client side, this information is rebuilt to resemble the original environment. Another example is the web-based telerobotic system for research and education at Essex by Yu et al [7], which aims for enabling remote users to control and/or program their mobile robot to explore their laboratory. To do so, the system sends continuous 24-bit JPEG images to users as a feedback. To program a robot, a user have to submit a text-based script, based on the Colbert language, containing commands for the robot. After the program uploaded, a Colbert evaluator is run to analyse the script. However, based on the analysis on our current system, we hypothesize that there are several aspects needed for making the remotely-accessible robotic laboratory more bene cial for users in general and students in particular. First is interoperability with robotic simulator. While experiment with real robots is an ideal and can bring more satisfaction to users, however it takes time and resources. Moreover, it is hard to debug. Thus, we see the necessity of using a robotic simulator. Even though we know that codes that run perfectly in simulator will not always run in real robots, but to have a seamless integration between the simulator and the remote laboratory might reduce the development time and e ort during the experiment. Furthermore, by allocating proportionally the load between simulator and real robots, we can have higher e ciency and utilization of robots. To this respect, we are studying the feasibility on integrating some robotic simulators to our Teleworkbench; some of them are Pyro [1], Microsoft Robotic Studio [4], and Webots [2]. Second is the need of analysis tool to assess the behaviour of the robots during experiments. When the program running on the robot, it is di cult to see what is actually going on inside the robot because we can only see the behaviour from outsider perspective. However, in this way, we might wonder why a program runs perfectly at times but not at some other time. Thus, we need somehow a tool to see inside of the robots during runtime. So far, we have partially achieve this requirement. We have developed a post-experiment analysis tool using video based on MPEG-4 standard [5]. Third is enough feedback and interactivity. Without enough feedback, users will nd it di cult to interact with robots. And without good interaction, they will easily get bored. Moreover, if the experiment requires situation assessment from users, e.g. telerobotics with supervised control, it must be very hard to do it properly. Thus, we need to transfer the remote environment where the real experiment runs to the users as much as possible. One way