Collision Avoidance Systems (CAS)

Introduction:
UAS is an emerging industry in both commercial and military/government use. The implications of drones has been seen as immensely valuable and can save companies and enormous amounts of time and money. As drones began to develop in the early 21st century, many made sure to note the possible safety concerns and hazards that follow alongside autonomy. In response, collision avoidance systems have been developed and implemented to ensure that accidents occur far less than what was possible previously. CAS consist of two key components; sense/detect and collision avoidance. The details of each component as well as specific systems will be described below

Sense/Detect: 
Unmanned aircraft will be put into operations where conditions can change or unknowns will appear. These conditions were not planned for and therefore the platform is unaware of how to handle such situations. The key component of sensing and detecting possible collisions is to ensure the UAS can quickly gain knowledge of the possible danger and act at almost the same exact moment. There are multiple methods in which an aircraft can acknowledge possible threats and collisions.

(Figure.1)

• Visual Sensors - These sensors generally come in the form of a camera that allows the aircraft to extract information regarding the obstacle or environment and process that information quickly. Much like human vision, these sensors are able to visually see the obstacle, relay the information to the main control (in this case our brain) and that in return acts accordingly. Visual sensors are often light and small and can be easily attached to a drone without sacrificing other aspects of the aircraft. There are some downsides to these cameras, however. Since these sensors rely on visually identifying objects and obstacles, they are susceptible environmental factors. Light, weather, and colors can all negatively affect the performance of these cameras much like they can all effect our vision. These cameras can sometimes easily mistaken dust particles on the lens as a hazard. On top of all this, their inability to determine distance can also be a large downside. However, this form of detection is deemed quite reliable and is used in many modern operations. (Figure.1)

• Radar - This is another form of detection that is found on many modern vehicles that require sensing the environment. Radar can quickly scan areas and relay that information back to the control. Radar also does not require the object or obstacle to be directly in line of sight which can prove to give more positive results. This type of detection is also not as susceptible to environmental factors like visual sensors. However, radar is not used on unmanned aircraft due to its sheer size which drones are not able to withstand or contain.

Collision Detection:

Detecting a collision is the next step in CAS. Once the UAS receives the information it has obtained from the detection system, it must have the capabilities to know how to avoid it. This includes knowing aspects about the obstacle and turning that into movement or an action. This component also needs to understand if the object is even an obstacle that it must avoid and must be able to understand its surroundings to determine if CAS is possible. Below are a few possible collision detection types.

(Figure.2 Trajectory Calculation on OctoCopter)

• Trajectory Calculation - (Figure.2) This type of collision detection takes into account the distance the obstacle is from the aircraft. The aircraft is set with a specific threshold that, if it dips underneath that threshold, will take action to avoid the object. 
• Worst Case Estimation - This method determines the trajectory of the obstacle and lays out every possible trajectory it could possible have and choose the one most suited for the situation. This is the most reliable but not used because it is also quite inefficient.
• Probabilistic Estimation - This method requires an immense amount of computational power and is inefficient because of it. It takes the probabilities of each path the aircraft could take and determines which one would be the most efficient.
• Act As Seen - This essentially works hand in hand with visual sensors. Whatever the camera picks up as information and relays to the computer is used to determine possible obstacles. Whatever is visually available to the drone will be determined if it is a hazard.

Conclusion:

When it comes to collision avoidance systems there are many different types and components that come into play. Depending on the type of operation and the specific platform that will be flown will help to determine what methods are best. Regardless of the method, collision avoidance systems help keep aircraft and others safe out of harms way. In today's society where autonomy is growing rapidly, perfecting CAS is crucial.