Unmanned aerial vehicles (UAV) or drones have developed very fast in recent years, and they have started to be applied in many fields such as aerial photography, military, and law enforcement. In a conceivable future, people will deploy more drones for civil and commercial uses which are not fully implemented today. One example is parcel delivery. Drones in large numbers will carry parcels to different destinations. To limit the movement of those drones. One idea is to construct a ”drone road,” a virtual tube-like area in the airspace. Drones should usually cruise within the tube but be able to move outside to avoid collision, or other hazards. One of the biggest problems, similar to ground self-driving cars, are collisions. Collisions will cause severe economic loss and even threaten people’s lives. Therefore, collision avoidance for drones is a valuable and important topic to research.

We propose a coordination-free algorithm for drones to avoid collisions on the drone road and present the result of a simulation-based analysis to evaluate how some parameters related to wireless communications can affect performance. This study first proposes a system model to clarify the boundaries of the collision-avoiding problem we attempt to solve. In this model, drones are required to fly within a long straight drone road with the same forward direction. Drones are only equipped with a GPS sensor and wire- less communication components to perceive the environment. Moreover, only position and speed information can be shared with other drones. Then we propose a criteria-based algorithm specifying when drones should switch to another lane to avoid an incoming collision. Otherwise, drones should slow down their speed to wait for a good chance. Eight criteria are put forward in this research, including two baseline criteria ”do nothing” and ”always slow down.” Other parameters considered in simulations include beacon interval, path loss exponent for log-distance model, transmit power, and drone density. We run simulations with those parameters on OMNet++ for evaluation. We furthermore present a cost model to represent the performance to quantify simulation results.

The simulation results show that packet loss is the main reason for collisions in our scenario. All simulation parameters can severely affect the packet loss rate and then consequent the collision rate. Four out of eight criteria can significantly reduce the total cost compared to the baseline.