Unmanned Systems Technology 010 | nuTonomy driverless taxi | Embedded computing | HFE International marine powertrain | Space vehicles | Performance monitoring | Commercial UAV Show Asia report

28 Unmanned vehicle digest | nuTonomy driverless taxi taxi’s proximity to the ground limits the radar’s line of sight). “Vision, on the other hand, gives you much richer information up to the horizon, but often lacks high- precision depth of information because we use a monocular imager.” Cameras The imager is configured to look in front of the nuTonomy taxi in order to detect and track relevant objects, monitor traffic lights and maintain a general record of the journey, and for this purpose the company chose the Blackfly camera from Point Grey Research. It is installed inside the front windshield, with the electronics running through the length of the chassis roof to the rear storage compartment of the taxi, where the processing hardware is installed. Although nuTonomy declined to disclose precise specifications, the Blackfly is capable of recording traffic imagery at up to 5 MP in resolution and at frame rates up to 90 fps. Versions are offered with either CMOS active pixel sensors or CCDs, supplied by Sony or Sharp, depending on model variations and requirements for resistance to light and noise. The camera’s common architecture includes a 16 Mbyte frame buffer, LED status indicators, and on-camera image processors for colour interpolation, gamma, and look-up table operation. In addition to the Blackfly cameras, a 1080p USB webcam from Logitech is installed by the driver’s seat at about eye level to enable logging of events that occur during vehicle operation, including negotiations of intersections and interactions with pedestrians. Radar A single, forward-facing radar system is installed for detecting vehicles in the surrounding area. As is often the case with unmanned automobiles – and conventional cars using radar for adaptive cruise control or collision mitigation systems – the radar is mounted behind the front bumper, invisible to onlookers and protected from possible impacts. Although it was originally integrated externally, this was changed during development. Unlike other driverless road vehicles, the nuTonomy Zoe does not rely on information from a rear-facing radar. “Such a sensor is currently not used, as we obtain a sufficient data set for our model using other types of sensors.” Iagnemma explained. Rather than following earlier car radar technology by using multiple beam radars with mechanical scanning or multiple fixed, overlapping beams, the system fitted on the taxis, the Electronically Scanning Radar 2.5 from Delphi Automotive, is intended to provide efficient coverage using a single radar, providing the vehicle with two simultaneous measurement modes: the first gives a ±45 º field of view (FOV) at a mid-range of 60 m; the second extends to 174 m with a ±10 º FOV. The radar’s mid-range mode detects road vehicles attempting to cut in from adjacent lanes, and picks up pedestrians and additional vehicles across the width of the Zoe. The long-range mode provides data on the approach speed, angle and range of vehicles further ahead, with object discrimination algorithms that allow for up to 64 targets in the taxi’s path to be identified. The dual functionality allows support for collision warning as well as adaptive cruise control, headway alert and mitigation, and brake support. Independent of mode, the radar operates at 76.55-76.78 GHz, has an update rate of 50 ms and a radial velocity from -100 to 25 m/s. The additional use of Simultaneous Transmit and Receive Pulse Doppler waveform allows for independent measurements of range and radial velocity, and detection of clustered stationary objects. Lidar Of all the systems involved in the nuTonomy Zoe’s sensor architecture, it is Lidar technology that is used most extensively during operation, while also providing potentially the greatest quality and quantity of navigation and safety- critical data. “The maximum safe speed of the vehicle is closely related to the range of the Lidar sensors, since if the perception range of the vehicle is less than the vehicle’s ‘reaction distance’, there is a heightened risk of collision,” Iagnemma says. October/November 2016 | Unmanned Systems Technology The taxi uses corner-mounted Lidars near the ground to detect pedestrians and obstacles