October 5, 2018 | Written by: Dan Cunnington
Categorized: Innovation | Research
Share this post:
Supported by the Hartree Centre’s Innovation Return on Research programme – Emerging Technology and colleagues from the STFC Hartree Centre have been working to contextualise the driving environment, fusing multiple data sources together to allow vehicles to make intelligent decisions. Following on from a recent publication at the IEEE Vehicular Technology Conference, the team recently presented an Extended Abstract at the ACM Computer Science in Cars Symposium in Munich, Germany titled “A Cognitive Assistant for Route Selection Using Knowledge Heuristics”. This utilises cognitive load weights learned from the driver in order to dynamically provide routing capability to vehicles that avoids areas of high cognitive load for the driver.
We utilised data from an experiment performed by Schneegaas and his colleagues in Stuttgart, Germany to identify areas of high, medium and low cognitive load for different participants in the experiment. For example, the following image shows areas of low cognitive load in green, medium cognitive load in yellow and high cognitive load in red for one of the participants.
Using techniques proposed in our earlier IEEE VTC paper, we performed a static feature extraction of OpenStreetMap data for the Stuttgart region to identify different road types and road features:
This allowed us to generate a set of cognitive load weights to inform the vehicle which road features caused the greatest cognitive load for each of the participants in Schneegass’ experiment. For example, here are the cognitive load weights shown on a scale between use and avoid for a participant in Schneegass’ experiment:
We then built an A* search over the OpenStreetMap graph where the cost function is dynamically calculated based not only on the distance to the destination node, but also on the perceived cognitive load of the particular road segment. Here you can see a route chosen that navigates in a fairly direct manner as this participant isn’t stressed by any of the roads between the start and destination. When the weights are altered to those of another participant, you can see the route changes accordingly:
As you can see, this participant doesn’t like residential roads so therefore receives a route that avoids residential roads.
We envisage dynamic routing being an important aspect of autonomous vehicles, as they attempt to navigate their environment, whilst taking into account user preferences, such as if you would like to take a call, or if the vehicle cannot go into autonomous mode, which route you may prefer to drive.