Focus on Electronics

---

April 2002
More 1 2 3 4

The driver distraction dilemma

Driving on a flat, straight highway during daylight hours when the weather is wonderful may not sound like a particularly dangerous driving situation; however, a majority of accidents occur under these exact circumstances, according to Barry Kantowitz, Director, Univ. of Michigan Transportation Research Institute (UMTRI), and a participant in the "State of the Industry for Driver Focus Research" panel presented by Motorola at the SAE 2002 World Congress.

"Accidents happen when the world appears okay," said Kantowitz. Psychological causes for driver distraction include "switching," when mental attention is working on one process and not another, and "starvation," which means there is insufficient mental processing capacity at a particular moment. He also pondered what an "intelligent" vehicle might want to know in order to keep operating. Some of these criteria are vehicle parameters such as velocity and lean position; telematic demand (collision avoidance, entertainment and convenience functions, etc.); local environment such as traffic; and driver characteristics including age, alcohol detection, hours of service, and whether the driver is eating.

"We should consider the option of a completely autonomous car [that can] operate independent of the driver" to help curtail driver distraction, said Kantowitz. He also presented various ways of removing distractions, including warning labels and government regulations on such things as cell-phone use, but concluded that "the best solution is good engineering design. If you build it right to begin with, then you won't have driver distraction; no telematic device would add to driver distraction. And where does good engineering design come from? It comes from good engineering judgment, and that comes from making mistakes."

The session also highlighted some of the goals of and research deriving from the Motorola Research Summit held in August 2001. Some areas for which answers were sought include the effects of various in-vehicle, secondary tasks on drivers; how an intelligent agent could be used to manage information flow to focus drivers' attention; and the characteristics of cell-phone users and their patterns while driving. Paul Green of UMTRI detailed a current research project focusing on cell-phone users. The research should be completed this spring, said Green.

Other panelists included John Lee of the Univ. of Iowa and Mike Gardner, Director of Intelligent Systems Research Lab, Motorola.

- Ryan Gehm

Motorola engineers simulate CAN, LIN communications

Engineers at the Motorola Virtual Garage of Detroit can model and simulate a whole-vehicle network.

With more and more electrical components and computation power being integrated into vehicles, a communication network is needed to reduce the complexity of the vehicle wiring harness, increase reliability of the overall system, and enhance the flexibility of the entire system layout. Controller Area Network (CAN) is one type of in-vehicle network standard developed by European OEMs in the past 10 years and has been adapted by U.S. OEMs in the past two years. Integration of a cost-effective, reliable CAN network can be a difficult task. Discovering system-level communication network problems, identifying ECU impacts at early development stages, and gaining more confidence of the CAN network behaviors are all critical goals, according to Motorola engineers who spoke at the SAE 2002 In-Vehicle Networks technical session. Vehicle-level CAN network modeling and simulation is one way to start answering these problems.

Engineers at the Motorola Virtual Garage of Detroit modeled and simulated a whole-vehicle CAN network and selected eArchitect from Innoveda to set up the vehicle network simulation environment. For demonstration purposes, they constructed a vehicle-level CAN communication strategy, which is captured in the Vector's CAN database format.

This communication strategy includes three CAN sub-networks (CAN B, CAN C, and CAN Diagnostic) and requires gateway functionality to share signals between different CAN sub-networks.

The models used in the simulation were developed using the EML language used in eArchitect and captured in the library format. Modeling vehicle communication protocols is the key for vehicle network simulation. The capability of modeling networks including CAN, LIN, MOST, J1850, and TTP is the heart of vehicle network simulation. In this simulation, researchers implemented CAN communication protocols based on CAN specification 2.0-B.

The outputs of simulation provide rich sets of information regarding vehicle network behaviors, node behaviors, and communication task behaviors. In addition to the eArchitect analysis tool, the post process tool developed by the Motorola Virtual Garage of Detroit provides a convenient method to analyze the simulation results further.

The use of CPU and throughputs of the communication bus are important for a message strategy design. With this simulation environment and the analysis tool, simulation results can provide detailed insight before the actual vehicle network is built. System engineers will be able to use this information to define a CAN message strategy intelligently and be able to control system integration and system latency time better.

The CPU performance of each node also plays important roles to maximize CPU use and guarantee the required performance. This simulation environment provides information on the CPU load due to the communication tasks. With a proper OS model (such as OSEK) and task scheduling running on the OS, this setup can also be used to analyze the performance of an OS and the efficiency of the scheduler.

This simulation environment has been used to conduct several studies of CAN networks, LIN networks, and mixed CAN/LIN networks to validate message strategies. The simulation environment also has been validated through hardware setup with good correlations.

The future for this simulation environment will include introducing OS into the CPU model and more communication protocol models. Introducing OS will improve modeling of CPU performance, which will lay down the foundations for the simulations of a distributed system.

- Linda Trego

More 1 2 3 4