Orchestrating Robot Swarms with Java

Tell us a bit about the work that you do.

I've been in the software industry for just over 10 years now, working primarily as a software developer and team lead. I have worked mostly in Java across several industries, telecommunications, finance, investment, banking, pensions, and now in automation. I've seen lots of different business, and all the software problems tend to be the same. I benefited from that, from my experience of different pressures either writing code for your business, writing code for users, B2B, B2C. I've been seeing a broad range of the software development industry. Recently I actually have moved into a more strategic role a bit away from day-to-day coding in the last maybe five or so months, getting into a bit more about a high level strategy at the company, having an input into the strategic decisions about what products we're building and what software we're developing and a bit less focused on how we are developing it.

Your talk is "Orchestrating robot swarms with Java." The robot swarm is totally run by Java?

In large part yes. We use the analogy of air traffic control, the control system which is telling all the robots where to go, what to do, making the most efficient use of them. That is run in Java.

How does that affect your system? How do you deal with GC, with pauses?

We are using plain old Java, very few frameworks. In our particular app, and the one we talking about, no Spring, no frameworks, not running on a web server. Just running as a JAR, containerised now. Very standard in that sense. However, the app is far from a standard app. Yeah, latency is a serious challenge in our application. And so naturally through our use of Java we have invested considerable effort in optimising garbage collection. And I'll be talking about some of that in the talk. But because we're dealing with real hardware, with warehouses which take many months to build and get to their full capacity. I'll also mention the heavy use of simulation techniques and how that helps us to validate new designs, new facilities, that they can achieve the desired throughput. How we use simulation to evaluate new algorithms when we're looking to optimize. And then some of the other implications that real-time control systems throw up as well and how we tackle that.

What are some of the takeaways from your talk?

I hope the audience will begin to consider some concerns which they may have never thought about before, to broaden their perspective of what Java can do, and a bit about where some of the pitfalls are and how you can get around them.