Behind the Test Challenges of Automotive Radar Systems

Transcription

1 Virtuelle Instrumente in der Praxis VIP 2017 Kurzfassung Behind the Test Challenges of Automotive Radar Systems David A. Hall National Instruments Corporation, Austin, USA Vor mehr als 50 Jahren, nämlich bereits 1959, verbaute man im Konzeptauto Cadillac Cyclone XP-74 zwei abgewandelte und urspünglich aus der Luftfahrt stammende Radarsysteme, die den Fahrer vor Verkehrsaufkommen warnen sollten. Heute ist ein Radar im Fahrzeug kleiner als ein Eishockey-Puck und wurde vom Konzept zur Realität. Einfachere Systeme zur Vermeidung von Auffahrunfällen und unerwünschten Spurwechseln werden mittlerweile durch Fahrerassistenzsysteme (Advanced Driver Assistance Systems, ADAS) ersetzt, wodurch sich auch neue Herausforderungen bei der Entwicklung und dem Test der Systeme ergeben. Moderne ADAS-Architekturen kombinieren komplexe Sensorik, Verarbeitungs- und algorithmische Technologien miteinander und bilden das, was im Endeffekt das Innenleben der autonomen Fahrzeuge ausmacht. Für den Verbraucher bedeutet der gesteigerte Einsatz von ADAS-Technologien ein Plus an Komfort und Bequemlichkeit. Für Ingenieure jedoch bringt die Evolution von ADAS zum einen sichere Arbeitsplätze mit sich, zum anderen aber auch unterschiedlichste Herausforderungen in Systementwicklung und -test. Abstract More than 50 years ago in 1959 the Cadillac Cyclone XP-74 concept car featured two modified aircraft radars that were designed to alert the driver for the presence of oncoming traffic. Today, an automotive radar sensor is smaller than a hockey puck and has transitioned from concept to reality (fig. 1). Figure 1: Nose cones on the front of a Cadillac Cyclone concept car housed modified aircraft radars. 238

2 Behind the Test Challenges of Automotive Radar Systems As primitive systems for collision and lane change avoidance are being replaced with advanced driver assistance systems (ADAS) they introduce new design and test challenges. Modern ADAS architectures combine complex sensing, processing, and algorithmic technologies into the what will ultimately become the guts of autonomous vehicles. As consumers, the growth in ADAS technology provides comfort and convenience. As engineers, the evolution of ADAS provides a combination of job security mixed with very different design and test challenges. State of ADAS Technology As ADAS systems evolve from simple collision avoidance systems to level 5 autonomous vehicles, they require increasingly complex sensing and computing technologies. For example, consider the sensing technology on the Tesla Model S, which combines information from 8 cameras and 12 ultrasonic sensors as part of its Autopilot technology. In addition, numerous reports have pointed to the 2018 Audi A8 as the first vehicle capable of Level 3 autonomy using radar, camera, and LIDAR sensors. Adding to the complexity, autonomous vehicles like these are increasingly using sensor fusion to combine multiple sensor inputs from radar, camera, ultrasound, LIDAR to better interpret objects and obstacles. This approach takes advantage of the benefits of radar for range detection and the benefits image processing algorithms for identifying what those objects are. Although sensor fusion can significantly improve the accuracy with which an autonomous vehicle detects obstacles it has significant implications on the processing capabilities of autonomous vehicles. Figure 2: Sensor fusion utilizes decision-making using inputs from multiple types of sensors. As a result, autonomous vehicles utilize significantly more complex processing technologies and generate more data than ever before (fig. 2). As an example, the Tesla Model S actually contains 62 microprocessors more than three times the number of moving parts in the vehicle. In addition, Intel recently estimated that tomorrow s autonomous vehicles will produce 4 Terabytes of data every second. Making sense of all of this data is a significant challenge and engineers have experimented with everything from simple PID loops to deep neural networks to improve autonomous navigation. 239

3 David A. Hall Importance of System-Level Test The combination of increasingly complex ADAS technology from sensor fusion to artificial intelligence algorithms demands a new approach to system design and test. For example, one can imagine the challenges of testing the performance of an unpredictable algorithm like a deep neural network. Engineers testing automotive radar systems are no longer answering the question of does the radar tell the right range?. Instead, they are asking can my sensor fusion algorithm tell the difference between a pedestrian and a road sign? Therefore it is no longer enough just to test the physical signal (transmit/receive RADAR signal). Instead, test techniques like object simulation are hard requirements. Answering such questions are essential for the development of autonomous driving and uncovering those answers requires more than a simple voltage measurement or reading a CAN packet. Instead, it requires engineers to recreate the embedded algorithm s physical environment with hardware-in-the-loop (HIL) test systems. These systems run vehicle models in real time and provide fake signals to one or more of a vehicle s electronic control modules (ECM) in order to convince the ECM that it is operating under real-world conditions. Typical HIL systems use automotive networks such as CAN, FlexRay, and BroadR-Reach as well as electrical signals with built-in fault injection to simulate a vehicle s interaction with the ECM. As the impetus to apply the brakes shifts from a person pushing a physical pedal to a computer sorting through terabytes of data HIL test systems must also evolve. Today, HIL test systems must accommodate technologies like ADAS and V2X communication and must re-create new environmental scenarios like the electronic signature of a pedestrian crossing the street. Figure 3: Typical radar test patterns Additionally, multiple sub-systems must now be tested together to ensure that the entire system functions properly. It is no longer enough to merely test the radar system, camera system, and brake system separately. Instead, engineers are increasingly required to test the entire chain to make sure the vehicle will actually stop when an object is detected. As a result, HIL test solutions are evolving in a manner that allows them to simulate multiple sensors types using tightly synchronized instrumentation, bus interface, and data acquisition modules (fig. 3). 240

4 Behind the Test Challenges of Automotive Radar Systems NI s Platform-Based Approach to HIL To address these challenges, NI offers a platform-based approach to hardware-in-the-loop testing. Using HIL test techniques, engineers are able to rapidly develop and test highly complex control systems by simulating the physical environment. Key features of NI s HIL test approach is the tight synchronization between a wide range of PXI instruments which enables engineers to simulate many unique driving scenarios and sensors. Particularly important in ADAS and autonomous driving applications, NI s FPGA technology enables engineers to design HIL test systems with extremely fast loop rates in order to test quick decision making. One recent example of an HIL test system that utilizes NI s platform-based approach to sensor fusion testing was demonstrated by the ADAS Innovations in Test ( consortium at NIWeek 2017 in Austin, Texas. This group is a joint collaboration between NI alliance partners S.E.T., Konrad Technologies, measx, and S.E.A. At NIWeek, the group demonstrated a comprehensive ADAS test solution which is able to synchronously simulate radar, LIDAR, communication, and camera signals for an ADAS sensor. In Figure 4, observe that the solution was able to simulate a virtual test drive using the IPG CarMaker and NI VeriStand software (fig. 4). Figure 4: Simulating a virtual test drive using IPG CarMaker Software, NI VeriStand Software, and PXI modular hardware In parallel with simulating the physical signals that correspond to the simulation, modular systems such as PXI simulate many of the physical signals effectively recreating the physical environment of the ADAS sensor or ECU. Note in Figure 5 that synchronization is a critical requirement of the PXI modules because all of these signals have to be precisely simulated in parallel, as the ECU needs the radar, V2X and camera signal to arrive at the same time in a real-world vehicle to process and understand the scenario and act accordingly. 241

5 David A. Hall Figure 5: Test Configuration for ADAS ECM Using HIL test techniques based on a platform, engineers are able to simulate a virtually unlimited duration of driving time and achieve greater test coverage to better understand how the embedded software performs in a wide range of situations. As a result of simulating driving conditions in the lab environment, engineers are able to identify critical design flaws much earlier in the design process. For example, at Audi AG, the radar team recently adopted a PXI-based system radar simulation. According to project lead Niels Koch, radar hardware-in-the-loop simulation enabled them to simulate ten years of sensor environments within few weeks. Going forward, as sensors continue to escalate in complexity HIL testing is an absolute requirement as a method to validate virtually unlimited driving time very quickly and therefore reduce time to market (fig. 5). Summary Fifty years ago, a front-mounted radar system that notified the driver of oncoming traffic was a gimmick. Five years from now, it will be the difference between life and death for passengers in autonomous vehicles. Given the impending safety and regulatory considerations of this technology, engineers will utilize HIL test techniques to literally simulate billions of miles of driving. Advanced systems like autonomous vehicles are quickly re-writing the rules for how test and measurement equipment vendors must design instrumentation. In the past, test software was merely a mechanism to communicate a measurement result or measure a voltage. Going forward, test software is the technology that allows engineers to construct increasingly complex measurement systems capable of characterizing everything from the simplest RF component to comprehensive autonomous vehicle simulation. As a result, software remains a key investment area for test equipment vendors and the ability to differentiate products with software will ultimately define the winners and losers in the industry. 242