Rodney Brooks | Hacker News Reader

Rodney Brooks

technologyreview.com

36 pointsprostoalex11 comments

Rod Brooks was on my thesis committee. His work was inspirational, but the subsumption architecture was badly flawed and ultimately proved to be a dead end. It's simply not true that "the world is its own best model" because wherever you are you can only see a small part of the world. It works great if all you want to do is avoid obstacles, but if you want to do anything complex and goal-directed, a robot really does have to be able to remember things. The reason he was so influential is because the dominant paradigm at the time, the so-called sense-plan-act model, was just as much of a dead end, and on top of that, produced really boring demos so it was doubly doomed. The right answer is to combine the two approaches with a layer of "glue" that connects them together in a so-called "three-layer" architecture [1]. That's actually the way that autonomous robots work today (AFAIK -- I've been out of the business for a very long time now).

[1] http://www.flownet.com/gat/papers/tla.pdf

My first job out of school nearly 30 years ago was building commercial research autonomous robots. We used a subsumption arch for real time avoidance and path following linked to a reactive planner implemented in C and xlisp for goal directed actions and task execution. While most modern books on robotics skim the history of this area - there are a lot of good ideas from that period of research that can be cheaply revistited. If you know anyone interested in robotics - get them to read Ron's (lisper) three layer architecture paper - its a classic in terms of summarizing previous approaches and it also that provided a foundation for a lot what now drives today's research.

re: NeilV - Braitenberg "Vehicles" book - enough goodness cannot be said about that book. It seems deceptively simple but its a wonderful place to start thinking about and being inspired to building simple beasties.

As should Grey walter's wonderful tortoise: http://www.bristol.ac.uk/news/2008/212017945378.html

I wonder under which paradigm the modern robots of iRobot (the Roomba company, where Brooks works today) are developed. Them being three-layer would be a pretty good indictment of pure-subsumption as an architecture.

> iRobot (the Roomba company, where Brooks works today)

That was two companies ago, AFAIK

Edit: And, indeed, TFA says as much.

TFA?

The Fine Article

I quite appreciated Rodney's paper from 1985 titled "A Robust Layered Control System for a Mobile Robot" [1]. I'm sometimes wondering if mobile robotics would look much different without this paper or if we would arrive to something similar anyway.

I even asked Rodney Brooks whether the results of this paper were applied on some robot platform as a follow up research and received a very interesting reply [2]. I agree that I might have worded it poorly but it's still one of my favourite tweets.

[1] https://apps.dtic.mil/dtic/tr/fulltext/u2/a160833.pdf [2] https://twitter.com/WeeklyRobotics/status/105775034299205222...

Great guy. Here's an anecdote from that time. He spent a few months on sabbatical at our lab ( https://ai.vub.ac.be/ ) in those days. We were preparing robots for a NATO Advanced Study Institute ( https://www.springer.com/gp/book/9783642796319 ), and I was struggling writing the serial driver for a custom embedded computer for this as the system kept crashing (due to a bug in the dram controller). Anyways, Rod Brooks, offered to help me with the coding. It wasn't needed as the code was not the problem, but I don't know many professors that could and would be prepared to dive in that deep.

P.S. I posted this comment in reply to a Brooks reference a few days back, but might be more on topic here.

A useful search term the article didn't mention: https://en.wikipedia.org/wiki/Subsumption_architecture

Before you get into that, it's delightful to start with the Braitenberg "Vehicles" book: https://en.wikipedia.org/wiki/Braitenberg_vehicle

I love how he's constantly dubious about autonomous robots news on twitter. Yesterday he criticized the Berkeley delivery https://twitter.com/rodneyabrooks/status/1165736688141787136... for not being autonomous and their short distances. He's also very skeptical about self-driving cars in general.

Allen sounds remarkably like a new version of Grey Walter's Machina speculatrix: https://sites.google.com/view/machinaspeculatrix/home

Reply on news.ycombinator.com