- 上一篇：柴油机连杆英文文献和中文翻译
- 下一篇：PLC工业机器人英文文献和中文翻译
agents near the obstacles.
Although the work of Olfati Saber addresses the problem of flocking using a simple point mass agent with double integrator dynamics, the results are sup- ported by a sound theoretical analysis. It will be interesting to see if the same analysis can be extended to more dynamically complex vehicles. It is of course clear that extending this idea to highly nonlinear vehicles (e.g. helicopters) will constitute a big challenge in this respect. Samilo¨gluGazi et al. [13] give an ex- ample of how a simple physical constraint like a restriction on the turn angle may lead to oscillatory behaviours in the group.
2 The idea
As we have clearly seen in the introduction, achieving the flocking or swarming of real vehicles with complex dynamics is still an unsolved problem. Our work addresses many of the issues involved in this area: we aim to build a flock of dynamically complex vehicles (i.e. microhelicopters) to perform flocking in a real world scenario where the dynamics of the vehicles and the noisy outputs of the sensors are not negligible. The use of an aerial robotic platform removes the two dimensional limitation to which most of the previous research has been constrained, allowing for a scenario more similar to the one normally experienced by fish or birds.
In order to reduce development time and research costs, we aim to leverage as much as possible of the technology available in the market place - in other words, to take a COTS (commercial-off-the-shelf) approach. This translates into select-
ing a suitable commercially available vehicle (see section 3) and fitting it with the necessary off-the-shelf components; however, some hardware will inevitably have to be designed (see section 3.1).
These helicopters are of course structurally identical, but differences in the electric motors, the trim of the blades and the swashplate mechanism, and de- formation of the very flexible foam blades will result in quantitatively different dynamic properties. The design of our controller should take this variability into account, together with the changes that will be induced by different sensor in- strumentations of the same helicopter. These considerations mean that it will be more appropriate to develop a general method for automated model identi- fication and controller design that can then be applied to different inpidual helicopters with different dynamics. A method based on machine learning tech- niques and artificial evolution is proposed in 4.
3 The helicopter platform
The ability to move in three dimensions is deemed to be an essential requirement of our system, as well as the need to be usable indoors (for ease of development). Only a few platforms can fulfil those two constraints: lighter than air vehicles (e.g. small blimps), and miniature helicopters. Small aircraft and slowflyers are clearly not an option, since our research arena (a cylinder of 12m diameter and 6m height) is too small to accommodate a flock of them. The more favourable size to payload ratio when compared to blimps led us to settle for a rotary wing solution.
Every roboticist and aircraft model enthusiast knows that the lift and hover- ing capability typical of a helicopter come at the cost of reduced dynamic stabil- ity. The helicopter flight controller is therefore a key element of our system. After evaluating several other models, a helicopter with a counter rotating dual rotor configuration was chosen for this study [14] (see figure 1). The counter rotating configuration is well known for delivering high efficiency as well as achieving excellent stability thanks to the direct compensation of the torque between the two rotors. The model we have selected uses a conventional fully controlled lower rotor, and an upper rotor fitted with a 45 degree stabilising bar. The stabilising bar exploits gyroscopic forces in order to counteract sudden changes in shaft in- clination. This results in improved stability, but of course this comes at the cost of a reduced response to control commands. Testing by human pilots showed the model to be more stable and easier to fly than conventional single rotor heli- copters; although not suitable for advanced aerobatic manoeuvres, the helicopter retains the manoeuvrability necessary to perform flocking.