Biomimicry, detection and information processing: shaping the intelligent systems of the future.

Across industries, especially in mobility, the trend is toward empowerment and connectivity. Infrastructure, furniture, equipment, household appliances, vehicles, planes, drones, trains or metros: systems must be ever more intelligent, connected, responsive and reliable.

Depending on the industry and the context considered, they must be able to detect and interpret an environment, understand or anticipate an ongoing phenomenon or measure physical data with precision, including in the event of partial failure.

Manufacturers, whoever they may be, are therefore developping all kinds of sensors, particularly optical or reflective.

And for the information collected to have value, it must obviously be processed and induced an adequate and rapid response from the system and its calculation units, particularly for embedded systems in the mobility sector.

Signal processing methods, the algorithms that govern them, memory management and calculation speed are quickly becoming technological issues. These systems and intelligences will also be the subject of intense competition.

How can we increase the multisensorality and reliability of systems and develop new types of sensors? How to improve detection skills? How to develop fast and simple artificial intelligence?

Information management was not invented by man. Nature is much more experienced and can offer many particularly interesting lessons.

Nature orchestrates complex ecosystems. To interact, species must understand the environment around them and therefore analyse the information they perceive. Hearing, sight, touch, sensitivity to magnetic fields, chemoreceptors: the intelligence of perception methods is almost infinite.

Processing complex signals, correlating them and making appropriate decisions is the most common activity in all living species. Each individual has a nervous system and a brain that allows them to respond to complex, sometimes unprecedented, situations to ensure their daily survival. The processing of information is not an animal monopoly, since it is also carried out in the plant world.

Natural treatment principles and systems are highly effective, regardless of the neuronal capacities of the species studied. They are both simple and sophisticated, extremely efficient and fundamentally sober in their design and consumption.

A specialist in the field, Bioxegy explains to you why and how biomimicry is a source of solutions and new approaches for designing the detection and information processing systems of the future. Here we provide you with a selection of particularly evocative examples.

The human vision: optimising the reaction speed of autonomous embedded systems.

Traditional methods of processing information involve analysing the perceived image, pixel by pixel. Their reactivity remains relative.

These techniques can be replaced by an approach inspired by the human “retina-cortex” binary. It invites us to no longer study the entirety of each image but to record modifications of certain global characteristics such as hue, speed or direction. This type of processing makes it possible to quickly eliminate irrelevant elements such as buildings or trees on the side of the road or an object moving away from the vehicle.

Inspired by human vision, the French company Prophesee has designed vision systems offering machines with reactivity that is close to ours. This technology is based on the observation of events rather than images.

More energy efficient and multifunctional, this bio-inspired system finds applications in many sectors: from automobiles to health, including robotics and industrial assembly.

Prophesee has a string of international awards: the company was nominated for the Silicon 60 Class of the American magazine EE Times for the third consecutive year in 2019 as well as winning the Top Innovation 2019 prize from inVISION magazine.

Crédits images : © Prophesee

Dragonflies: optimise anticipation of trajectories and detection of obstacles.

In the field of mobility, autonomous vehicles, trains, or drones must be able to understand their environment. To do this, algorithms must be able to analyse the reality of a given situation. Intelligent, they must be able to anticipate the trajectories of other objects.

In the case of land mobility: other cars, pedestrians, cyclists for example. Even if it is a faculty mastered by human beings, we are far from being the most gifted in this area. Many insects are much more reactive and clever than we are. They could inspire new methods of developping algorithms.

The dragonfly is a formidable predator: its success rate is estimated at 95% (compared to 50% for the white shark). It has 360° vision and an exceptional reaction time of 30 milliseconds. It is especially impressive with its ability to follow its prey, notably flies, midges and mosquitoes, and to anticipate their trajectories to better catch them in mid-flight.

This biological model of the dragonfly turns out to be less complex to study than that of humans and the still unknown mechanisms of their brain. Researchers from the universities of Lund in Sweden and Adelaide in Australia therefore looked at the insect to understand its ability to detect movements and anticipate trajectories in order to replicate it within artificial neural networks.

They designed an autonomous robot to test their innovation which would optimise the analysis of the trajectories of objects tracked in the vehicle environment. This technology will open the way to a better reaction to incidents and risky situations and the possibility of creating better maneuvers to avoid obstacles or other vehicles.

Crédits images : © University of Adelaide

Locust swarms: an unrivaled anti-collision system.

When it comes to computers and on-board systems, the question of overload often arises. How to prioritise information in a dense data flow? This is particularly the case in the field of mobility, for which it is important to develop rapid analysis systems capable of detecting collisions.

The prioritisation of perceived information is a specialty of nature, which must solicit the nervous systems of animals as little as possible to guarantee low energy consumption.
Here too, biomimicry can provide intelligent answers!

Let's look at the locust: this insect has the particularity of being able to move in swarms. Some clouds of locusts even reach more than 40 million individuals per km²! Despite this incredible density, the risk of collision between individuals is extremely low. It is clear that the locust's ability to avoid its congeners is extremely perfected and very rationaliqed, as the insect must react quickly by means of a nervous system that is infinitely simple compared to the human brain.

How do they avoid the collisions? The locust has a giant neuron inside its brain, responsible for controlling the insect.

The visual stimuli perceived by the locust generate electrical potentials travelling along the optic nerve to the giant neuron. This acts as a filter: for the vision of a locust, objects or congeners flying on a collision course generate a greater electrical potential than congeners flying in parallel.

These strong signals inducing higher voltage manage to pass through the giant neuron to reach the brain. Other signals are filtered and neglected. Thus, the locust is only concerned with a small number of elements and obstacles. The others are ignored automatically.

By studying the functioning of this neuron, researchers at the University of Lincoln have developped a fast and economical sensor for autonomous vehicles. Particularly useful research for developping algorithms capable of acting in traffic-laden environments or quite simply for de-saturating embedded system computers.

Crédits images : ©Bilal Tarabey / AFP

The elephant fish: ultra-sophisticated detection in disturbbed environments thanks to the electric field.

The elephant fish (Gnathonemus petersii) is a nocturnal fish found in African rivers. It evolves in particularly murky environments, turbulent and dark waters.

Despite everything, it is quite capable of spotting its prey and its peers. Its detection method is based on an electric field that it produces by muscular contraction. It then perceives the slightest variation in this electric field thanks to its electro-receptors to locate its prey or its peers.

Inspired by this electrical sense, the French company Elwave was able to develop a set of sensors allowing the detection and electromagnetic characterisation in real time of all types of objects at 360°.

Already used to guide underwater robots or detect buried objects, this technology can considerably reduce the risk of blind spots and poor visibility for a vehicle in circulation.

Crédits images : © Elwave

The nictitating membranes of camelids: ensuring the cleaning of sensors

There is little point in developping high-performance and sophisticated sensors if they become clogged and blind in poor weather conditions. Here is an example that Bioxegy worked on, in the automotive sector.

Current cleaning systems, which are nozzles similar to windshield nozzles, often require a lot of water and require very regular maintenance, which is impossible to carry out over long journeys. Consumption has even been estimated at 100 L/h in snowy weather!

Faced with this problem with one of its clients, Bioxegy turned to camelids, camels and dromedaries.

These desert mammals must confront the aridity of their environment and face particularly intense sandstorms. They still manage to clean their eyes of impurities and grains of sand.

They owe this very particular ability to their nictitating membranes, transparent third eyelids which ensure the protection of the eyeball as well as its cleaning. This "biological system" is particularly effective because it minimises the use of the tear glands which lubricate the cleaning action.

Inspired by this principle, Bioxegy's teams were able to imagine a mechatronic cleaning device for the optical sensors of autonomous vehicles, saving water and energy!

Read our full file

Crédits images : © Bioxegy

Desert ants: orient themselves in space using only light rays

Can we consider that a vehicle or train whose orientation relies exclusively on a battery of satellites as truly autonomous?

This is the question that arises for autonomous vehicles whose guidance is done by GPS. Indeed, this system presents certain limitations, apart from the quantity of infrastructure necessary for its operation, such as its resolution varying from 5 to 15 meters or its sensitivity to storms or fog.

The democratisation of autonomous means of transport, particularly in certain regions of the world, could quickly mean that a particular vehicle must find its way by itself without using GPS. But, what if the answer came from ants?

Desert ants (Cataglyphis) have an exceptional ability to position themselves in space. They are able to find their way without resorting to phenomena, these chemical receptors normally used by other species of ants.

Their positionning technique is based on a triple strategy: they study the polarizstion of sunlight, measure the distance by optical flow and count their steps. Researchers from the National Center of Scientific Research and the University of Aix-Marseille, at the Institute of Movement Sciences - Étienne Jules Mayer (ISM) have designed the Antbot robot, directly inspired by these desert ants.

Antbot replicates the navigation stratagem of ants and is able to return to its starting point after a random exploration of 14 meters within one centimeter. Note here that this technology requires very low computing capacity, therefore considerable savings in equipment, weight and energy. This technique could inspire other navigation technologies in many areas, notably mobility, and lead to simple and efficient equipment.