AI News, System prevents speedy drones from crashing in unfamiliar areas

System prevents speedy drones from crashing in unfamiliar areas

Now MIT researchers have developed a trajectory-planning model that helps drones fly at high speeds through previously unexplored areas, while staying safe.

The model—aptly named 'FASTER'—estimates the quickest possible path from a starting point to a destination point across all areas the drone can and can't see, with no regard for safety.

But, as the drone flies, the model continuously logs collision-free 'back-up' paths that slightly deviate from that fast flight path.

If, as we move along this fastest path, we discover there's a problem, we need to have a backup plan,' says Jesus Tordesillas, a graduate student in the Department of Aeronautics and Astronautics (AeroAstro) and first author on a paper describing the model being presented at next month's International Conference on Intelligent Robots and Systems.

In forest simulations, where a virtual drone navigates around cylinders representing trees, FASTER-powered drones safely completed flight paths about two times quicker than traditional models.

In real-life tests, FASTER-powered drones maneuvering around cardboard boxes in a large room achieved speeds of 7.8 meters per second.

As the drone flies, each detected voxel gets labeled as 'free-known space,' unoccupied by objects, and 'occupied-known space,' which contains objects.

To do so, 'convex decomposition,' a technique that breaks down complex models into discrete components, generates overlapping polyhedrons that model those three areas in an environment.

Somewhere along the whole trajectory, it plots a 'rescue' point that indicates the last moment a drone can detour to unobstructed free-known space, based on its speed and other factors.

That was computationally expensive and slowed down the drone's planning, so the researchers developed a method to quickly compute fixed times for all the intervals along the trajectories, which simplified computations.

'How to increase the flight speed and maintain safety is one of the hardest problems for drone's motion planning,' says Sikang Liu, a software engineer at Waymo, formerly Google's self-driving car project, and an expert in trajectory-planning algorithms.


Instead of controlling the drone to precisely follow the teaching path, our method converts an arbitrary jerky human-piloted trajectory to a topologically equivalent one, which is guaranteed to be safe, smooth, and kinodynamically feasible with an expected aggressiveness.

The proposed method is integrated into a complete quadrotor system and is validated to perform aggressive flights in challenging indoor and outdoor environments.For details we refer readers to our paper.

We have advanced our Teach-and-Repeat system as a complete and robust Teach-Repeat-Replan system which can deal with changing environments, moving obstacls and imperfect localization, and with all necessary components for an autonomous drone (global/local perception, global/local planning, global/local state estimation and onboard controller).

Unmanned aerial vehicle

The flight of UAVs may operate with various degrees of autonomy: either under remote control by a human operator or autonomously by onboard computers.[3]

The term drone, more widely used by the public, was coined in reference to the early remotely-flown target aircraft used for practice firing of a battleship's guns, and the term was first used with the 1920s Fairey Queen and 1930's de Havilland Queen Bee target aircraft.

A similar term is an unmanned-aircraft vehicle system (UAVS), remotely piloted aerial vehicle (RPAV), remotely piloted aircraft system (RPAS).[10]

UAV is defined as a 'powered, aerial vehicle that does not carry a human operator, uses aerodynamic forces to provide vehicle lift, can fly autonomously or be piloted remotely, can be expendable or recoverable, and can carry a lethal or nonlethal payload'.[11]

however, due to the wind changing after launch, most of the balloons missed their target, and some drifted back over Austrian lines and the launching ship Vulcano.[19][20][21]

After the 1973 Yom Kippur war, a few key people from the team that developed this early UAV joined a small startup company that aimed to develop UAVs into a commercial product, eventually purchased by Tadiran and leading to the development of the first Israeli UVA.[32][pages needed]

The images and radar decoys provided by these UAVs helped Israel to completely neutralize the Syrian air defenses at the start of the 1982 Lebanon War, resulting in no pilots downed.[39]

The first time UAVs were used as proof-of-concept of super-agility post-stall controlled flight in combat-flight simulations involved tailless, stealth technology-based, three-dimensional thrust vectoring flight control, jet-steering UAVs in Israel in 1987.[40]

The U.S. Military UAV tier system is used by military planners to designate the various individual aircraft elements in an overall usage plan.

Small civilian UAVs have no life-critical systems, and can thus be built out of lighter but less sturdy materials and shapes, and can use less robustly tested electronic control systems.

Also, properly designed, the thrust to weight ratio for an electric or gasoline motor driving a propeller can hover or climb vertically.

UAV computing capability followed the advances of computing technology, beginning with analog controls and evolving into microcontrollers, then system-on-a-chip (SOC) and single-board computers (SBC).

Degrees of freedom (DOF) refers to both the amount and quality of sensors on board: 6 DOF implies 3-axis gyroscopes and accelerometers (a typical inertial measurement unit – 

UAV actuators include digital electronic speed controllers (which control the RPM of the motors) linked to motors/engines and propellers, servomotors (for planes and helicopters mostly), weapons, payload actuators, LEDs and speakers.

These bi-directional narrowband radio links carried command and control (C&C) and telemetry data about the status of aircraft systems to the remote operator.

So instead of having 2 separate links for C&C, telemetry and video traffic, a broadband link is used to carry all types of data on a single radio link.

Advanced autonomy calls for situational awareness, knowledge about the environment surrounding the aircraft from exterioceptive sensors: sensor fusion integrates information from multiple sensors.[51]

for flight control) tick as fast as 32,000 times per second, while higher-level loops may cycle once per second.

Hierarchical control system types range from simple scripts to finite state machines, behavior trees and hierarchical task planners.

The most common control mechanism used in these layers is the PID controller which can be used to achieve hover for a quadcopter by using data from the IMU to calculate precise inputs for the electronic speed controllers and motors.[citation needed]

Medium levels of autonomy, such as reactive autonomy and high levels using cognitive autonomy, have already been achieved to some extent and are very active research fields.

Reactive autonomy, such as collective flight, real-time collision avoidance, wall following and corridor centring, relies on telecommunication and situational awareness provided by range sensors: optic flow,[71]

In September 2013, the chief of the US Air Combat Command stated that current UAVs were 'useless in a contested environment' unless crewed aircraft were there to protect them.[167] A 2012 Congressional Research Service (CRS) report speculated that in the future, UAVs may be able to perform tasks beyond intelligence, surveillance, reconnaissance and strikes;

the CRS report listed air-to-air combat ('a more difficult future task') as possible future undertakings.[168] The Department of Defense's Unmanned Systems Integrated Roadmap FY2013-2038 foresees a more important place for UAVs in combat.[169] Issues include extended capabilities, human-UAV interaction, managing increased information flux, increased autonomy and developing UAV-specific munitions.[169] DARPA's project of systems of systems,[79]

Israel companies mainly focus on small surveillance UAV system and by quantity of drones, Israel exported 60.7% (2014) of UAV on the market while the United States export 23.9% (2014);

Chinese drone manufacturer DJI alone has 75% of civilian-market share in 2017 with $11 billion forecast global sales in 2020.[84]

2018 NPD point to consumers increasingly purchasing drones with more advanced features with 33 percent growth in both the $500+ and $1000+ market segments.[86]

should rise from a few hundred million dollars on research and development in 2018 to $4 billion by 2028 and $30 billion by 2036.[93]

As global food production demand grows exponentially, resources are depleted, farmland is reduced, and agricultural labor is increasingly in short supply, there is an urgent need for more convenient and smarter agricultural solutions than traditional methods, and the agricultural drone and robotics industry is expected to make progress.[94]

Research is exploring miniature optic-flow sensors, called ocellis, mimicking the compound insect eyes formed from multiple facets, which can transmit data to neuromorphic chips able to treat optic flow as well as light intensity discrepancies.

Because of their small size, low weight, low vibration and high power to weight ratio, Wankel rotary engines are used in many large UAVs.

Micro air vehicles endurance is so far best achieved with flapping-wing UAVs, followed by planes and multirotors standing last, due to lower Reynolds number.[51]

Solar-powered atmospheric satellites ('atmosats') designed for operating at altitudes exceeding 20 km (12 miles, or 60,000 feet) for as long as five years could potentially perform duties more economically and with more versatility than low earth orbit satellites.

Another application for a high endurance UAV would be to 'stare' at a battlefield for a long interval (ARGUS-IS, Gorgon Stare, Integrated Sensor Is Structure) to record events that could then be played backwards to track battlefield activities.

Besides, dynamic assessment of flight envelope allows damage-resilient UAVs, using non-linear analysis with ad-hoc designed loops or neural networks.[113]

UAVs can threaten airspace security in numerous ways, including unintentional collisions or other interference with other aircraft, deliberate attacks or by distracting pilots or flight controllers.

although there was no significant damage to the balloon nor any injuries to its 3 occupants, the balloon pilot reported the incident to the NTSB, stating that 'I hope this incident helps create a conversation of respect for nature, the airspace, and rules and regulations'.[118]

Rogers stated in an interview to A&T 'There is a big debate out there at the moment about what the best way is to counter these small UAVs, whether they are used by hobbyists causing a bit of a nuisance or in a more sinister manner by a terrorist actor'.[120]

Several security researchers have made public some vulnerabilities in commercial UAVs, in some cases even providing full source code or tools to reproduce their attacks.[125]

At a workshop on UAVs and privacy in October 2016, researchers from the Federal Trade Commission showed they were able to hack into three different consumer quadcopters and noted that UAV manufacturers can make their UAVs more secure by the basic security measures of encrypting the Wi-Fi signal and adding password protection.[126]

In 2016, Transport Canada proposed the implementation of new regulations that would require all UAVs over 250 grams to be registered and insured and that operators would be required to be a minimum age and pass an exam in order to get a license.[133]

The ENAC (Ente Nazionale per l'Aviazione Civile), that is, the Italian Civil Aviation Authority for technical regulation, certification, supervision and control in the field of civil aviation, issued on 31 May 2016 a very detailed regulation for all UAV, determining which types of vehicles can be used, where, for which purposes, and who can control them.

In 2015, Civil Aviation Bureau in Japan announced that 'UA/Drone' (refers to any airplane, rotorcraft, glider or airship which cannot accommodate any person on board and can be remotely or automatically piloted) should (A) not fly near or above airports, (B) not fly over 150 meter above ground/water surface, (C) not fly over urban area and suburb (so only rural area is allowed.) UA/drone should be operated manually and at Visual Line of Sight (VLOS) and so on.

'Hobby drones' with a weight of less than 7 kg at altitudes up to 500m with restricted visual line-of-sight below the height of the highest obstacle within 300m of the UAV are allowed.

In July 2018, it became illegal to fly a UAV over 400 feet (120 m) and to fly within 1 kilometre (0.62 mi) of aircraft, airports and airfields.

On 21 June 2016, the Federal Aviation Administration announced regulations for commercial operation of small UAS craft (sUAS), those between 0.55 and 55 pounds (about 250 gm to 25 kg) including payload.

Certification of this position, available to any citizen at least 16 years of age, is obtained solely by passing a written test and then submitting an application.

For those holding a sport pilot license or higher, and with a current flight review, a rule-specific exam can be taken at no charge online at the website.

Commercial operation is restricted to daylight, line-of-sight, under 100 mph, under 400 feet, and Class G airspace only, and may not fly over people or be operated from a moving vehicle.[155]

Additionally, CNN's waiver for UAVs modified for injury prevention to fly over people, while other waivers allow night flying with special lighting, or non-line-of-sight operations for agriculture or railroad track inspection.[158]

In Oregon, law enforcement is allowed to operate non-weaponized drones without a warrant if there is enough reason to believe that the current environment poses imminent danger to which the drone can acquire information or assist individuals.

Putting the Power of a Film Director in an Autonomous Drone

Commercial drone products can tackle some automated tasks, but one thing those systems don't address is filming artistically.

It autonomously understands the context of the scene — where obstacles are, where actors are — and it actively reasons about which viewpoints are going to make a more visually interesting scene.

As a goal, 'artistically interesting' is subjective and difficult to mathematically quantify, so the system was trained using a technique called deep reinforcement learning.

For example, other autonomous drone products often use a continuous backshot because it allows the drone to follow a clear, safe path behind the actor.

While the system averaged users' preferences for shots as an actor walked a narrow corridor between buildings, it can apply those preferences to similar obstacles like a forest path using topographic mapping.

'Future work could explore many different parameters or create customized artistic preferences based on a director's style or genre,' said Sebastian Scherer, an associate research professor in the Robotics Institute.

Other innovations include efficient motion planners to anticipate the trajectories of actors, and an incremental and efficient mapping system of the environment using LiDAR.