Unmanned Systems Technology 025 | iXblue DriX I Maintenance I UGVs I IDEX 2019 I Planck Aero Shearwater I Sky Power hybrid system I Delph Dynamics RH4 I GCSs I StreetDrone Twizy I Oceanology Americas 2019

87 because they do not want to take original manufacturers’ systems outside of their validation specifications. “The simplest way for us to interact with the powertrain is to use the existing pedal interface, requiring no reverse engineering or ‘hacking’,” O’Sullivan notes. The StreetDrone Twizy has a maximum endurance of about 64 km (depending on fitted equipment) and a top speed of 80 kph. Between three and four hours of runtime can be delivered by the system’s battery during typical operations, which fits the profile of the Twizy as a vehicle for two- to three-hour runs. The Twizy takes four hours to fully charge, and is designed for Level 1 chargers, at 120 V AC. “We have some third-party adapters that let you use Level 2 charging but they use the same current as the household one, so it’s a convenience thing – it won’t actually charge it quicker,” says O’Sullivan. Up to 300 W of power is delivered via a DC-DC converter in the trunk, where the self-driving computation systems – typically Nvidia Drive PX2 platforms and their counterparts – are installed. O’Sullivan says, “We have designed our own mount that lets you connect up the self-driving hardware and take it out easily, as well as all the interfaces for the cameras and sensors to be integrated with minimal effort.” The company is also in the process of integrating a back-up battery for the trunk-based electronics, for running the computational systems while the Twizy is switched off. Although most of the base powertrain has been left untouched, StreetDrone needed to fit its own actuators and drive- by-wire system controllers in order to make the Twizy autonomy-capable. Fortunately, the architectural simplicity of the Twizy, combined with a scheme by Renault to share the Twizy’s CAD and CAN databases with companies such as StreetDrone, made the vehicle an ideal proving ground for this type of system redesign. “We’re not limited to using the Twizy, as there’s no hacking of the CAN system and there’s nothing in our approach that is vehicle-specific, so the same approach we use can be applied to other vehicles,” O’Sullivan says. The base Twizy platform lacked a power steering motor to begin with, so the company fitted a control unit to actuate the steering, as well as “stress- tested mountings, our own motor on the steering column, and a motor controller attached to the main control unit,” O’Sullivan says. “That gives us a very simple CAN interface where we can request steering angles in software.” The braking system also had to be redesigned so that it can be actuated by a motor, which required significant use of CAD software, testing and installation of performance monitoring sensors to track their behaviour. The sensors in particular were critical for ensuring the safety case of the drivetrain, for example to identify a motor current that’s inconsistent with a steering or braking request, which could mean the driver being handed back control of the vehicle until the error causing it can be addressed. That is especially helpful for the bulk of users, who are from universities and other academic or programming backgrounds, and who want to focus on their algorithms rather than spending months on developing and validating a functional, safe drive-by-wire system from the ground up. O’Sullivan adds, “Our very first Twizy, which was delivered to a customer a year ago, had full control of the acceleration, throttle pedal, steering and the brakes. We’ve since developed and integrated into our vehicles the means to control the horn, indicators and the lights, and to monitor the reverse lights and brake pedal, which are all things that the [UK] Department for Transport asks for when you’re performing public testing on roads.” In addition to the safety driver, the Twizy comes with a second seat where an engineer can sit and monitor the various data coming in from the car’s sensors for discrepancies between the data and performance targets. Sensor architecture Naturally, the drive-by-wire system depends on a suite of sensors to feed it data inputs, which it can use to output actions as determined by the self- driving algorithms. For monitoring vehicle dynamics, the wheels integrate MEMS speed sensors, which are designed as wheel-like devices with concentric ‘teeth’ that piezoelectrically detect the number of teeth passing in StreetDrone Twizy | Digest Unmanned Systems Technology | April/May 2019 The StreetDrone Twizy is designed for typically two- to three-hour operations on test tracks and city roads, but can operate for up to four hours depending on use