STW.nl

U bent hier

P14-18 i-CAVE, the integrated cooperative automated vehicle

With the presentation to the public of the automated vehicle by Google, huge awareness was created on the capabilities and possibilities of automated driving.

Automated driving, such as done by Google and others, can however not solve our everyday mobility problems. Doing this will require both automation and cooperation between vehicles and other (non-equipped) road users: what we need is cooperative automation. Our i-CAVE project has this direct focus and will allow for significant steps on the HTSM Automotive roadmap.

Cooperative driving has already been researched and piloted for a decade in many European projects. The SARTRE project, demonstrating a mixed platoon of trucks and passenger vehicles is the most well-known, but many other projects have been executed or are ongoing, such as the Connect&Drive project, the Grand Cooperative Driving Challenge, and many others. Whereas automated driving aims to automate individual vehicles based on situation awareness obtained by onboard sensors, cooperative driving utilizes a wireless communication link among vehicles and between vehicles and infrastructure. This wireless link can be used as an additional source of information, enabling automated functions, but usually still involves some control by the driver. The main benefit of wireless information is twofold: First, information can be obtained regarding other road users, which are not in the line-of-sight (as opposed to the use of onboard sensors); Second, information from other vehicles can be obtained that cannot be (accurately) measured otherwise, such as acceleration, brake pressure, and throttle position. As a result, cooperative driving allows to optimize the traffic system with respect to throughput, fuel consumption, and safety. So far, the main focus has been on cooperative longitudinal control (i.e., speed and lead vehicle distance). The next step is adding cooperative lateral control (i.e., steering). To do this safely, more involved levels of autonomy are required from the vehicle.
In the current state of the art, automated vehicles aim to fully automate the driver’s task. The automated vehicle, knowing its destination, finds its way through the road network, while passengers are doing other things (just like passengers in public transport). To this end, the automated vehicle is aware of its position on the infrastructure network and is aware of other traffic, including vulnerable road users. This awareness, however, is limited due to the line-of-sight characteristics of onboard sensors. Although the automated vehicle is demonstrated and Google is planning to make the next step by producing and operating a fleet of fully autonomous “pods”, there are significant technical and human factors challenges to overcome, before these vehicles will actually contribute to safety, efficiency and current traffic problems. Regardless of these technical challenges, the typical automated car does not optimize its actions with other vehicles, so as to jointly optimize the total utilization of the entire road network. This is however exactly what is required to solve our society’s mobility challenges. Therefore, automated driving requires more involved levels of cooperation.

Clearly, both cooperative driving and automated driving have come to a point that, in order to make significant progress and impact, both technology roadmaps need to be combined. This idea is what we propose in our i-CAVE project and is supported by a Technology assessment “Tem de robotauto” conducted by the Rathenau institute. In this assessment it is even stated that: “the autonomous car developments should support cooperative driving by integrating these autonomous technologies and capabilities into the cooperative driving roadmap”.

The i-CAVE project utilizes human factor and technological aspects which are needed in both automated (and autonomous) and cooperative vehicles. The key challenges are related to:
• Highly accurate and scalable digital maps and video-based sensing and positioning technologies;
• Individual cooperative vehicle control taking into account vehicle dynamics and tire behavior, including robust software technologies;
• Behavioral human factors issues for drivers, passengers, and other road users (such as non-equipped vehicles and vulnerable road users)
• Managing a fleet of cooperative autonomous vehicles in an integrated way through a road network

Maps
Key to next-generation cooperative automated driving is improving the vehicle’s ability to perceive its environment. The current state-of-the-art in automated driving is to incorporate all static objects, such as traffic lights, traffic signs, pedestrian crossings, sidewalks, etc. of the environment in highly accurate (centimeter level) maps. The automated vehicle accurately positions itself in these maps, using sensing technology, and thereby is able to navigate through the assumed to be static environment. In addition, the vehicle uses sensing technology to detect and track dynamic objects, such as other vehicles and pedestrians, in the environment. On basis of this dynamic information, the vehicle performs continuous path planning, in order to prevent dangerous situations and collisions.

There are three key drawbacks of this approach that prevent large-scale deployment of current automated driving technologies. Firstly, the creation of the required highly accurate maps is extremely complex and requires careful and time consuming human verification. Due to this, deployment of this approach at national scales is currently not socio-economically viable, let alone at international scales. Secondly, the map is only useful, when the vehicle is able to accurately position itself in this map. This requires specific sensing technologies, which are currently not robust to weather conditions, like rain, fog, snow. In such conditions, safe automated driving is currently not possible. Thirdly, the capability of the automated vehicle to perceive static objects that are not in the map is extremely limited. In a sense, the map is a digital virtual rail road over which the vehicle travels safely. When the map is not up to date, for example a temporarily traffic light has been installed, the vehicle will not perceive it, leading to extremely dangerous situations. It is fair to say that highly accurate maps are currently the enabler of autonomous driving but are also its Achilles heel. Fourthly, the automated vehicles do not take into account the compatibility with normal driving behaviour, being acceptable and predictable for road users and vulnerable road users, ensuring a safe and acceptable situation.

In i-CAVE, we take two intertwined approaches to tackle these sensing related challenges of automated cooperative driving. Firstly, we develop innovative self-learning computer vision technologies that allow the vehicle to better perceive its surroundings, thereby lowering its dependency on highly accurate maps. Secondly, we develop computer vision technologies that allow for increased automation of map creation, thereby improving the scalability of using highly accurate maps. By doing so, we bring closer nation-wide deployment of automated cooperative driving technologies.

Vehicle control and fleet management
A well-known application of cooperative driving is platooning. This particular application is commonly referred to as Cooperative Adaptive Cruise Control (CACC). CACC involves longitudinal vehicle automation (throttle and brake), which may be extended with lateral automation. Here, wireless inter vehicle communication allows for very short inter vehicle time gaps (around 0.3 s) without compromising safety. This, in turn, greatly improves road throughput and, especially for trucks, reduces fuel consumption. However, road traffic is not limited to platooning situations, but also involves maneuvers such as splitting from the platoon and merging into the platoon. Although a platoon of CACC vehicles should be robust against non-equipped vehicles cutting in, true optimization of traffic efficiency and safety is achieved when cooperative vehicles are able to automatically split from and merge into the platoon. This illustrates the necessity to combine the road maps associated with automated and cooperative driving, as mentioned before.
The implementation of so-called “cooperative and automated maneuvering” not only requires automation of the individual vehicles, but also automatic control of the interaction between vehicles. It is envisioned that wireless inter vehicle communications are the main enabler to perform such coordinated maneuvers. Note that highly accurate scalable digital maps are of the essence for this type of automation, since these maps allow vehicles to assess the region of interest and to predict the trajectory of other vehicles. In addition, in particular traffic situations, such as highway slip roads and crossings, the maneuvering can be supported by intelligent roadside units, capable of tracking the vehicles involved in the maneuver and wireless communication with cooperative vehicles.
Therefore, the second objective of i-CAVE is the design an experimental evaluation of distributed (i.e., vehicle-based) controllers for cooperative and automated maneuvering.

Human factors
The transition from fully manual driving to fully automated driving (level 5 as described by the SAE) is a very challenging one from a HF point of view. In case of level 5, at which the driver does not have to have any control over the vehicle, the driver becomes a passenger and main HF questions are restricted to acceptance, comfort, and safety. But in order to have full flexibility, dual mode cooperative vehicles are the right step forward. This means that when the conditions requires this, vehicles can be operated in a manual mode, but when technology is ready and cooperative technology gets introduced, we can shift to level 2 of SAE (the vehicle having lateral and longitudinal control, with the i-CAVE addition of cooperative technology) and even to level 3, 4 and 4 if we use this technology in the right circumstances. This means we have to configure a fail-safe driver-vehicle symbiosis in which the driver knows exactly what the vehicle will do under what circumstances, and that cooperative technology is primarily used under those conditions that it adds to throughput, traffic safety and comfort.

Besides research on the topics above, a small fleet of test and demonstrating vehicle will be setup and used for research and demonstration purposes. These will be probably small vehicles, an electric urban shopper, which will have more easy access for technology enhancements and human factors research.

The envisaged results of the project are innovation building blocks, which can be used in industrial valorization roadmaps. To realize this, we will bring together an industrial consortium.
The general results of the project are aimed for usage, implementation and roll out in vehicles using the public infrastructure. However, the step towards this goal takes longer than we have in the i-CAVE project, so we will focus as a first implementation step on realization on dedicated “closed” areas, like business parks, leisure sites or event sites, but with the bigger long term picture as ultimate goal. However, in the dedicated environment we are aiming for developing and experimenting for real life conditions, so the system allows sometimes that the driver takes full control, sometimes the vehicle takes over control, sometimes the vehicle drives empty autonomously to the next destination to pick up a passenger, in other cases a platoon of driverless vehicles with only a heading vehicle with driver is repositioning a handful of cars into the managed road network.

Initiatiefnemer(s) 
Prof dr ir Jan Bergmans, TU/e; Dr Gijs Dubbelman, TU/e; Prof dr ir Gabriel Lodewijks, TUD; Prof dr Marieke Martens, UT; Prof dr Henk Nijmeijer, TU/e; Prof dr ir Peter de With, TU/e;
Topsector 1 
HTSM (inclusief ICT, Nanotechnologie en Medische Technologie)
Roadmap/ Innovatieagenda (indien van toepassing) 
Automotive NL smart mobility research roadmap
Contactpersoon 
Dr.ir. Ben Rutten, TU/e
TU/e Strategic Research Area Smart Mobility Ben Rutten Den Dolech 2 5612 AZ Eindhoven ben.rutten@tue.nl