On March 24th 2023, Eloy completed the Zenzic CAM Scale-Up Testbed phase with a demonstration day at UTAC Millbrook. We wanted to share our path to this point and how Zenzic fits in with our long-term plans.
Multi-vehicle coordination and its applications
Eloy has been fascinated with developing methods to improve vehicle movement across the road network through in-vehicle interventions. The most obvious missing piece in how we manage our roads is that we lack sufficient mechanisms to improve vehicle coordination with direct driver instructions. This story starts almost 20 years ago during my postgraduate research into traffic modelling in unmanned vehicles:
As part of a postgraduate research project, I was tasked with devising mathematical approaches to modelling the formation of phantom traffic jams and creating ways that could limit their impact. Most of us will have experienced phantom traffic jams when a driver overbrakes due to a small incident or rubbernecking. The drivers behind then sequentially also overbrake due to human reaction times, and the larger the reaction the more they need to overbrake to maintain a suitable headroom between vehicles.
Prof Eddie Wilson explains how phantom traffic jams form
If we limit potential solutions to phantom traffic jams to outside the car, we become quite restricted in we can do. We can have variable speed limits, we can collect data, we can plan ahead, and we can introduce smart motorways.
Many of these tend to be large projects with high costs. They are also broadly homogeneous, in that we provide all vehicles with a single rule, as it is too difficult to display different messages to cars without becoming confusing.
During my project work, I investigated models where I provided each vehicle with a different instruction. In effect an in-vehicle message for each driver and vehicle to follow. Then by running a pattern recognition over the entire swarm of road vehicles, a different intervention could be deployed based on the likely traffic pattern that could then emerge.
A few examples included:
- Making interventions anywhere rather than at fixed points
- Avoiding over acceleration
- Breaking up platoons and higher density traffic with specific car messages
- Managing lane changing
- Providing a recommended speed as a dampener
With these additional actions the phantom traffic jam formation was reduced in severity. However, in 2004 we were faced with a different technical barrier: no iPhones and no ability to provide these direct interventions in-vehicle. Also, at the time self-driving vehicles were still a dream – the Mojave Trials were just about to start – so automated solutions were yet to be considered.
Fast forward to 2019:
Over the years we saw incremental improvements in the available technology. Key to Eloy was the evolution of mobile app technology with the release of Apple CarPlay and Android Auto. This not only allows drivers to choose their own in-vehicle technology but also allows a path into vehicles for new technology creators without having to battle through complex OEM supply chains.
Mobile phones are also really convenient for drivers. Being able to adjust their settings from the comfort of their sofa rather than in-vehicle meant that technology education could take place as well as familiarity and marketing.
So, the technology barriers I saw in 2004 have gone. The remaining barrier, which many technology innovators face, is how to get sufficient participation to obtain the network effect of technology. How do we achieve a high level of use at a specific time and location? The obvious answer: think about destinations where everyone is travelling based on a fixed start or end time.
One challenge with large destinations is the car parks – could we create specific in-vehicle interventions that could manage the car park capillaries at scale? There was a necessity to build an in-vehicle traffic signal system that could learn the car park layout quickly and without a high level of human input. An in-vehicle traffic signal has 2 clear distinctions between a physical roadside traffic signal: it can be far cheaper to deliver, and it can be tailored to each vehicle. This is where we see a clear difference in MVC from GLOSA (Green Light Optimized Speed Advisory or “Greenwaves”) which is simply adding a digital information system to an already existing physical system (so it isn’t cheaper and it can’t be tailored).
Our in-vehicle traffic signal system had stumped the Eloy team. How could we build a sequence of escalating tests without any evidence the technology would work? Then in late 2021 the Rees Jeffreys Road Fund offered a competition for new road technology – what could we build that might be relevant to our roads in 50 years. This felt like a great place to start.
Fellow applicant Phil Carey proposed a narrow rural roads traffic signal system that blended really well with Eloy’s vision of a connected car platform. We worked together on a final application and were fortunate enough to win! Understanding how rural roads and car parks have intrinsically linked needs for in-vehicle traffic signals might not be easy to spot at first but this will become clearer!
Using mobile phone technology to coordinate vehicles down narrow rural lanes
When we stuck our teeth into the rural lane challenge, we noted that a single pair of traffic signals might work for a short stretch of road with just a few passing places but what happens on a stretch of road with tens or hundreds of passing points and more than a few cars. Technically we saw a very strong overlap between the narrow rural road challenge and the car park challenge: we needed to be able to build an artificial intelligence system that could take in the road and route information (including road width) and self-determine the traffic signal locations and timing itself. The in-vehicle traffic signals for rural roads also needed to be cheaper to deliver and tailored for vehicles, such as caravans and HGVs.
Sequential traffic signals
With a narrow rural road or a car park consisting of many rows and paths we needed a way to coordinate the various virtual traffic signals we could create. Focusing on rural roads it was important to consider the flow of cars across the entire stretch of road and to be able to deliver a solution with just mobile phone technology so that it was cost effective.
There have been previous attempts at traffic signal coordination, with green waves tested all the way back during the Slough Experiments along with discussions on Cooperative Driving for Autonomous vehicles in SAE J3216.
Our preliminary work with the Rees Jeffreys Road Fund led us to present a 2-vehicle system at UTAC Millbrook as part of the Transport Technology Forum in September 2022.
Zenzic CAM scale-up
This brings us to the Zenzic CAM Scale-up. Whilst we were able to do internal testing with 2, 3 or 4 vehicles as an Eloy team we needed a way to scale to 20 or more vehicles and in a controlled environment. Whilst taking photos of HGVs travelling from the opposite direction helped describe the problem, we needed a good environment to test without external interference.
Using mobile phone technology to coordinate vehicles down narrow rural lanes
Zenzic provided further support for CAM Testbed days at UTAC Millbrook and HORIBA MIRA, where we used their 1-mile long city courses to create a narrow rural road setup with 4 or 5 narrow sections where cooperative driving (multi-vehicle coordination) was required.
Day 4 - demonstration day
This brings us to the final day of testing and where we start to loop back to our large destination product. In Day 3 we were able to show multi-vehicle coordination (MVC) improving journey times by 17% as head-on collision risk and traffic formation was significantly reduced.
Day 4 was more about starting to create a full service with partners that could be deployed at events and festivals in the near term. We will have further discussions on events in upcoming blogs too. Combining the strengths of existing services, such as temporary traffic equipment makes a lot of sense: digital services are new and needs to continue with validation. We also note that physical roadside equipment will always have attributes that digital services cannot achieve: physical barriers prevent cars from travelling and physical signs are usually easier to observe. We note that this is slightly different from our initial premise of in-vehicle only – however we note that some aspects of a solution will be mostly in-vehicle and some aspects will continue to be physical.
Therefore, for our Demonstration Day we partnered with leading traffic equipment supplier SRL Traffic Systems. Their variable message signs (VMS) would assist with communication to the professional test drivers, temporary (and solar powered) traffic signals would help provide instructions to the test drivers and we linked the VMS messages with Eloy’s own in-vehicle signage system (IVS).
Technically the demonstrations went very well. Driver interventions continued to work at the right moments, and we have confidence that the correct signals are getting to each vehicle in a timely manner. Similar to Day 3, we saw better journey times with the software running and almost no traffic jam formation on the narrow sections.
Our final demonstration at UTAC Millbrook
The additional aim of Day 4 was to share our work with guests from across the road transport sector, with a focus on those who deliver services for large destinations and venues. We gained important insight into how other organisations want our technology to iterate towards a market-ready event traffic management service.
The project and support from Rees Jeffreys Road Fund and Zenzic has been vital to getting our technology close to market ready. The initial version of the traffic software will be closer to our previous work on the RAC Veteran Car Run, which included providing drivers with a controlled and centrally managed route to the Eloy SatNav via a web portal. We will then work to incorporate the MVC technology so that we are no longer waiting for such long periods at these large destinations.
We are currently looking for pilot partners and destinations. Please contact Eloy if you have specific problems this technology can help solve.