Re: Routing and behavioural norms?

Hi Stefano,

Thanks for the interesting links, and best of luck with your paper! Robot forklifts are rather different from slime agents in respect to the complexity of the actions they are tasked to perform, so I am not sure how much they have in common from a technical perspective.  One of the challenges for forklifts is to decide whether and how to respond to requests for service, or whether they need to recharge their batteries by visiting a wireless charging station.  Requests are broadcast, and may be limited to agents within a limited distance. Message passing is thus one-to-one or one-to-many on a named topic.  Another example relates to collision avoidance. Mobile agents transmit information on their course and speed to other agents in their neighbourhood, analogous to the FLARM system for gliders and light aircraft, thus allowing agents to change course and/or speed to avoid collisions.

p.s. I was part of the Dagstuhl behavioural norms working group, and am now hoping to apply what we discussed to enabling agents to learn behavioural norms. For this I envisage message passing as above, as well as pheromone traces. I have had to rethink my approach after realising that A-star on a grid wasn’t going to be a satisfactory solution. I now want to use a more human-like approach that should be more computationally efficient as well as allowing for richer behaviours.

Best regards,
Dave

> On 1 Jun 2023, at 09:31, stefano.mariani@unimore.it wrote:
> 
>  
> Hi Dave, hi everybody, 
> Stefano speaking here, we met at Dagstuhl in the “learning” working group 😊
>  
> I dunno if it is of interest for you, hence apologies if it’s not, but I’m recently working on *learning to communicate** with pheromone (deposit ~ send a signal/message, following ~ receive the signal/message).
> For instance, a first work I submitted to ECAI 2023 (still waiting for acceptance or most probably rejection 😅) is about agents learning to communicate with pheromones to cluster together or scatter around (I can share the yet unpublished work privately if you are interested)
> What I already found out is that agents can learn with minimal information communication and coordination policies more efficient and effective than the ones usually hard-coded by design.
>  
> I think this learning approach may suit your use case and goal, but pardon me if that’s not the case.
> Here are some videos of the simulations with a brief explanation: https://www.youtube.com/playlist?list=PLu56TE55PP02uiL0aQ_N5Z5xHPFqEmrV_
>  
> Feel free to ping me if you want to know more or think I can help 😊
>                (the same holds for *anyone interested in this mailing list*, of course!)
> My best!
>  
> -----------------------------------------------------------------------------------
> Stefano Mariani, PhD
> Post-doc researcher @ University of Modena and Reggio Emilia
>     > https://smarianimore.github.io <https://smarianimore.github.io/>
> -----------------------------------------------------------------------------------
>  
> Da: Dave Raggett <dsr@w3.org> 
> Inviato: giovedì 1 giugno 2023 10:06
> A: public-webagents@w3.org
> Oggetto: Routing and behavioural norms?
>  
> I’ve been working on a web-based simulation of a swarm of robot forklifts in a warehouse:
>  
>                https://www.w3.org/Data/demos/chunks/warehouse/ <https://www.google.com/url?q=https://www.w3.org/Data/demos/chunks/warehouse/&source=gmail-imap&ust=1686211594000000&usg=AOvVaw0PQTEjRLg9Mb3KIgrDs5G9>
>  
> This work in progress, and the next step is to integrate automate route planning.  My initial approach has been to use the A-star algorithm across a grid of cells and to combine it with pheromone traces to enable collective learning.
>  
> The following web page randomises the origin and animates the search process. Use the retry button to select a new origin and initiate search for the destination:
>  
>                https://www.w3.org/Data/demos/chunks/warehouse/search/ <https://www.google.com/url?q=https://www.w3.org/Data/demos/chunks/warehouse/search/&source=gmail-imap&ust=1686211594000000&usg=AOvVaw2CLCY3SgL-yawjSOYfl0-9>
>  
> This shows that the A-star algorithm can provide far from optimal routes compared to what a human would devise. Some explanations for this include:
>  
> Search is essentially blind and feels its way around obstacles
> Limited beam width may result in promising routes being forgotten
> Routes may have kinks despite heuristics to avoid them
> Search finds the first solution, but further solutions may be better
>  
> A more promising approach is to treat the obstacles as defining a set of interconnected spaces with open areas, junctions and alley ways. In other words to convert the map of obstacles into a connectivity graph. The different spaces can then be associated with different heuristics, along with statistical priors and learnt behaviours. Pheromone traces provide one means to support a collective memory. I am working on a new demo to showcase how this can work.
>  
> A more ambitious goal will be to consider how to enable swarm agents to learn behavioural norms from their interaction with other agents, including ideas learned indirectly from other agents, courtesy of the hive mind.
>  
> Human drivers, for instance, learn when to give way to oncoming vehicles in narrow streets with parked cars as obstacles, and how to signal to each other via changes in speed and position, along with the turn indicator lights. At least in the UK, we also use our hands to thank the driver who gave way, and likewise, to acknowledge that gesture.  At night we flash the car’s headlights to say thanks.
>  
> Anyone interested in collaborating on this?  The warehouse is just one of many scenarios. Others include simulating urban traffic in cities, simulating smart factories, and logistics for container sea ports.
>  
> p.s. the underlying framework involves a collection of mobile and static agents that communicate using asynchronous messages. Cognitive control is based upon rules operating on knowledge graphs, inspired by John Anderson’s ACT-R architecture.
>  
> Best regards,
>  
> Dave Raggett <dsr@w3.org <mailto:dsr@w3.org>>

Dave Raggett <dsr@w3.org>