Figure 1. The GEROMS vehicle.
Résumé
Même la meilleure technique de pilotage d'un véhicule d'exploration planétaire est tributaire des communications avec la station sol. Plusieurs principes de communication ont été analysés pour des missions à destination de la Lune et de Mars. Des expériences reposant sur les performances types de quatre liaisons de transfert de données différentes ont été menées pour analyser la faisabilité d'un véhicule télécommandé et pour déterminer le degré d'autonomie requis à son bord.
Contractors:
ONERA/CERT/DERA (F) Eurogiciel (F)
Funding:
French national programme
Requirements on the communication link between a planetary rover vehicle and its control centre are mission-dependent and will differ amongst (for example) exploration missions to Mars, the rear side of the moon, or on the equatorial regions on the front of the Moon. The operation of a rover vehicle will also depend on the characteristics of the data link, ranging from direct remote control to full autonomy. It is thus important to match the mode(s) of operation to the data rate and time delays that can be expected for a particular mission.
To gain a better understanding of the factors involved, four different experiments have been performed to test various operational modes under realistic conditions.
For a mission to Mars, communications between a rover and the Earth operations centre can be direct or via an orbiter and the exact capacities of the data links between the rover and Earth are thus dependant on the specific features of a given mission. We can nevertheless assume that the links will only be available twice a day for perhaps one or two hours. The data rate may be as low as 1 or 2 kilobits per second for both the downlink and the uplink, accompanied by a long time delay of up to 40 minutes.
For a Moon mission, the delay time is about 2 seconds for a dedicated system during periods of direct visibility up to 10 seconds if a data relay satellite is used. Data flow may vary from a few tens of kilobytes up to 1 megabyte per second, but higher data rates incur severe penalties such as increased power consumption and stringent antenna pointing requirements.
The control of a vehicle in an unknown environment requires that the terrain to be crossed should be viewed and analysed in real time, either autonomously on board or by an operator on Earth.
Three-dimensional imaging using stereoscopic CCD cameras is the selected solution for terrain perception, since it involves purely static equipment which is easily qualified, and the techniques are now sufficiently fast and reliable to be used on a planetary rover.
Except when the vehicle is operated directly under remote control, its motion along a desired trajectory is directed by an on-board closed-loop controller, which uses inertial guidance techniques to cope with rapid changes in the contact between the wheels and the soil. The controller also ensures the safety of the robot.
When operated under direct remote control, images of the terrain in front of the vehicle are sent to the Earth station and displayed to the control operator. The effect of transmission delay on controllability has been the subject of experiments carried out with the vehicle developed by Geroms (Groupement pour les Essais en Robotique Mobile Spatiale). Stereoscopic terrain images and compressed video-data collected by panoramic cameras were displayed to a human operator who controlled the rotational speed, turning radius, pan and tilt angles of the camera. The time delay in the control loop could be adjusted up to 10 seconds.
Figure 1. The GEROMS vehicle.
A initial period was needed for the operator to learn to anticipate movements subject to large time delays, but thereafter, the vehicle was successfully driven, even over difficult terrain. The data in video images can be significantly reduced, by as much as a factor of 50, without noticeable impact on the ability of the operator to detect obstacles and select a safe path. The reulting data rate is typically 1 Mbit per second,
The advantage of remote control is a simple space segment. Its drawback stems from having to down-link a high data rate to Earth which requires a high gain steerable antenna, a high transmitter power and a primary power consumption of 100 W.
In order to reduce the refresh rate of the images, an on-ground computer constructed a digital terrain model (DTM) from stereoscopic image pairs. This was displayed to the operator who then drew the path to be followed by the rover. The path was then automatically adjusted to take account of the capabilities of the rover (such as its turning radius) and displayed back to the operator for confirmation. It was then sent to the rover which followed it under closed-loop on-board control. This mode was also tested on the Geroms vehicle.
The use of a DTM slows down the image refresh rate from ten images per second to one image per ten seconds. However the reduction in data is partially limited by the compression factor that can be applied to the stereoscopic image pairs since the stereo-algorithms are much more sensitive to the effects of data compression and decompression than a human operator. The complexity of the rover is nevertheless moderate.
A digital terrain model may also be generated in real time on board the vehicle, then transmitted to Earth. The operator then steers the vehicle whilst observing a three-dimensional representation of the environment seen from an artificial viewpoint ahead of the vehicle. The predicted view is synthesised from inputs provided by the operator and a mathematical model of the robot. The operator sends control variables to the rover which inputs them to a closed loop controller, to minimise the discrepancy between the model and the rover. Navigational variables are sent to the control centre at a higher rate than the DTM and are used to update the predicted position of the rover and cancel residual errors. This technique has been evaluated for one year on the experimental vehicle for exploration (EVE) and has proved to be robust.
Figure 2. The EVE test rover.
Generating a synthetic environment requires increased on-board computing power but is still viable, and a single PC601 processor ran all the necessary software. Although a DTM seems to be tolerant of data compression, the exact compression algorithms to be applied and the compression factors must still be investigated. However a 20- to 100-fold reduction in data transmitted to Earth (compared to simple remote control) is expected. An experiment to steer the rover from a remote site via a 64 kbit/s commercial data link is planned.
An autonomous navigation mode has been implemented in the EVE vehicle. The following sequence is used:
The longest path which will bring the rover to its goal, yet avoid dangerous and unknown areas is then computed, allowing for any uncertainties and safety margins. The computed trajectory is then followed. This cycle is repeated until the rover reaches its objective. The cycle can be executed in discrete steps or in a single manoeuvre during which the path is periodically updated. Evaluation of this mode for one year has been successful.
An automatic capability to determine the strategy of a manoeuvre has been added to improve the probability of successful navigation over hazardous terrains. The computing time needed to generate the trajectory is about 30 s compared to the 90 s needed to follow it. The data rate to Earth is limited to the monitoring of past events and is extremely low.
Four control modes and their operational requirements have been evaluated. These provide a repertoire of solutions to the mission constraints expected in the exploration of the Moon or Mars. Switching between two different control modes, travel under extreme conditions and scientific tasks, such as collection of samples, will be studied using the IARES vehicle presently under construction at CNES.
Preparing for the Future Vol. 7 No. 2