Issue 37 Unmanned Systems Technology April/May 2021 Einride next-gen Pod l Battery technology l Dive Technologies AUV-Kit l UGVs insight l Vanguard EFI/ETC vee twins l Icarus Swarms l Transponders l Sonobot 5 l IDEX 2021 report

30 April/May 2021 | Unmanned Systems Technology Dossier | Einride next-gen Pod unknown object is 75 m and 14 º to the right of the Pod, which may or may not confirm the measurement performed by its own sensor system. More external measurements can be received, either from other vehicles linked to the cloud- based system or from stationary sensors at strategic locations such as busy intersections, for example, or both. Additionally, the system can combine this information with intended states such as path predictions and planned routes for the Pod and all nearby connected vehicles, so that their intentions can be treated as independent sources of information. The system software can also carry out probabilistic filtration – with a bank of Extended Kalman Filter data fusion algorithms for example – to remove anomalies such as false negatives or positives from the retrieved information about the vehicle’s surroundings. Kalman filters are Bayesian mathematical processes that start with an estimate of a current state, such as a vehicle’s position and velocity, along with a degree of uncertainty, and then use new observations such as sensor data to improve the solution and reduce the uncertainty in a repeating cycle. The system can also transmit processed data packets with anomalies removed to other vehicles and/or the cloud after transforming them, so that they reference a global geographical system, a shared coordinate system common to a subgroup of the connected vehicles or a coordinate system local to the Pod itself. The company says it is this distributed system approach that brings redundancy and therefore greater accuracy to obstacle location while keeping the sensor suite required by each individual vehicle simpler and less costly than it would otherwise have to be to ensure accurate and reliable localisation and collision avoidance. The overall collision avoidance strategy that Einride has adopted, Hallgren explains, is to ensure that the system is at all times aware of any other actors in the environment around the Pod, so that it can adapt its behaviours to them. Along with the sensor fusion discussed above, the autonomy software uses a combination of environmental modelling and advanced motion planning to achieve this. While autonomous behaviours such as path planning and obstacle/collision avoidance are critical to safety, the base vehicles’ performance is also, naturally, important. In an emergency, for example, it can brake with a maximum deceleration of 0.5-0.8 g depending on the friction and grip available from the road surface, Hallgren says, adding that from 30 kph it stops in about 10 m. As a test and demonstration exercise, Einride took the Pod to a race track at Dunsfold Aerodrome in England in mid-October 2020, where it set a lap record for autonomous EVs, reaching a top speed of more than 80 kph and generating a maximum lateral acceleration of around 3.5 m/s 2 (0.36 g ). Driving at speed around the track showcased the capabilities of the sensing and path planning systems. Einride has published a video (go to www.youtube. com/watch?v=KQRuJU4EE4Q) based on data from them that shows short- and long-range Lidar scans of the track surface and edges, along with the racing lines through the corners worked out by the path planner and traced out ahead of the Pod as it drives its laps. In scoping vehicle performance Einride’s software provides intelligent transport planning and live data on emissions, energy usage, location data and more The overall collision avoidance strategy is to ensure that the system is at all times aware of any other actors in the environment around the Pod