A new Amsterdam Robotics meetup is being organised by Rajat Mani Thomas.
It was an interesting first evening. Here are some basic notes:
- Robot running time is expensive because of wear and tear. So machine learning (which takes time) is expensive.
- research is done in the lab: Limited State Space = very minimalistic environment where only a limited number of problems can occur.
- In Finite State Automata you have to write all the possible states, adding a new state means writing how it relates to all the others. A lot of work!
- In Fuzzy Logic each rule gets different scores, the most successful ones can be combined.
- In Optimal Action Discovery teams of robots experiment and establish possible actions and states. The system can learn to go from its current state to optimal state, through reinforced learning. Behavior patterns are pre-set, but have open ranges of change.
- The future for robotics: working in groups is as challenging for robots as for people. Object recognition is still difficult. (Color Invariant Object Recognition)
- ICRA is a mayor robotics conference: Rod Actuators, Minimal Hardware Design, visual markers on objects for the robot to be able to work (markerplants?)
- The real ‘ecology’ around robotics are teams who keep the system up and running, typically 5 PHD-s and 10 students. They tinker and then pray it keeps running for the experiment. Many of the videos you might see on youtube needed many runs to get a video of the robot working.
- The Botson Dynamics’ Bigdog is really the only robot that reliably works.
- So how do we get beyond the lab? The mechatronic part is okay, the AI is much more problematic.
- Robots are only starting to have a nervous system. The advantage of robotic systems is communication at near light-speed. The brain need not be central, but could be distributed. However communicational breakdown can ask for some basic processing power in the robot itself.
- What is becoming more apparent is that all learning systems are a function of the body. The process of learning is deeply embodied: to understand someone grabbing a cup of coffee, needs the same constraints mentally as grabbing it yourself.
- Healthcare as a market: based not on replacing but on assisting people in care. Or as a facilitator of communication.
- Academic field is communicating, but still very fragmented.