Thank you for writing such an insightful and lucid post!! As a robotics researcher, I’ve often found during implementation that hybrid approaches give me the best bang for my buck, but there’s so much excitement about pure end to end ML in robotics these days, that I often wonder if I’m just doing things wrong. It’s very interesting (and sobering!) to read about what you’ve found from having worked on this stuff for 8 years; perhaps hybrid systems deserve their own serious research agenda in addition to all the pure end to end ML :)
Did you see or anyone use old school Bayesian logic in robotics, like decision under uncertainty type (calculating utilities and such). I wondered whether they are useful or not for some type of problems.
Thank you for writing such an insightful and lucid post!! As a robotics researcher, I’ve often found during implementation that hybrid approaches give me the best bang for my buck, but there’s so much excitement about pure end to end ML in robotics these days, that I often wonder if I’m just doing things wrong. It’s very interesting (and sobering!) to read about what you’ve found from having worked on this stuff for 8 years; perhaps hybrid systems deserve their own serious research agenda in addition to all the pure end to end ML :)
Hi Benjie,
Did you see or anyone use old school Bayesian logic in robotics, like decision under uncertainty type (calculating utilities and such). I wondered whether they are useful or not for some type of problems.