This page contains project updates for CCSR.
Older videos:
Updates Jan 2nd '15:
- Changed voltage regulator to beefier 7A Pololu step-down regulator D15V70F5S3, the old 3A RioRand LM2596 couldn't handle the added arm-servos reliably.
- Added 4-DOF robotic arm
- Expanded object recognition and retrieval using robotic arm: CCSR can pick up objects based on color and size
- Expanded NLP capabilities; more sentence structures recognized, more lifelike behaviour using randomized responses.
- Added video showing some of the new capabilities
Updates Nov 27 '14:
- Added full sourcecode of CCSR application on github.
- Optimized NLP response speed: downsample audio to reduce filesize
- Gave nlpxCCSR access to CCSR hardware state: now you can ask verbal questions about how robot feels, what compass heading it has, how much battery it has left, etc (no video update yet)
- Expanded NLP capabilities; more sentence structures recognized, more lifelike behaviour using randomized responses
Updates November '14:
- Inspired by mtriplett, I added some basic speech-recognition, Natural Language Processing and Machine Learning capabilities to CCSR. Also added integration with cloud-based knowledge database (Wolfram Alpha). Implementation details follow below.
- Added video shows some initial experiments with using natual language to interact with CCSR
- Added software architecture diagram