D E SC RI PTI O N Technical field of the invention

The present invention relates to an electric scooter provided with innovative features in terms of safety and reliability. Among these, the scooter is provided with an anti-collision system and a method for managing and controlling the system. Specifically, the anti-collision system includes a virtual rear-view mirror for the scooter and a system for generating anti-collision alarms.

Applications of the anti-collision system, according to the present invention, may be in the field of micro-mobility vehicles, in particular electric scooters, in order to provide the driver with advanced functionalities typical of large vehicles. The control strategy of the system is intended to optimize driving by signaling possible dangers behind the driver. It is not intended as a 'safety critical' functionality as it does not interact in any way with driving.

Background art Electric scooters have been very popular, especially in cities, since the 2010s. Their success was determined by the development of electric motors light enough to be mounted on a conventional scooter, and more efficient batteries, such as lithium-ion batteries. They are considered means of micro-mobility, along with other similar means of transport such as

'segways', 'hoverboards' and 'monowheels'. Electric scooters generally have two small, hard wheels with a folding frame, usually made of aluminum. There is a platform between the two wheels for resting the feet, and the scooter is steered via a handlebar attached to the front wheel. Some scooters have three or four wheels, or are made of plastic, or are not foldable. The range generally varies from 5 to 50 km, and the maximum speed is about 30 km/h. Within the country, scooters, those equipped with an electric motor with power within 0.5 kW and a speed within 25 km/h, are equal to bicycles. They must have similar equipment to bicycles: two brakes, reflectors, front and rear lights (if absent, they may only be ridden during daylight hours), a horn or bell.

Like all means of transport, it has some criticalities due essentially to the type of urban mobility in which it moves. The main ones are as follows: a) the unpredictable behavior of drivers who, being able to move abruptly from a section of the carriageway occupied by cars to one occupied by pedestrians, cause serious difficulties for the drivers of other motor vehicles in predicting the direction of travel, also because of the unfamiliarity with which they have to anticipate changes of direction. It should be added that the road carriageway is not designed, if only minimally, for the circulation of these vehicles; b) the instability of the vehicle, which is often driven at higher speeds than bicycles and subjected to abrupt accelerations in an urban context whose narrow spaces in which pedestrians and other drivers circulate increase the danger; c) the severity of injuries. In the event of an accident, someone riding a scooter can be very badly injured even in the event of very slight contact because the balance of someone riding one scooter is quite unstable; d) frequent violations of minimum safety requirements such as carrying an additional passenger, failure to signal in low visibility situations.

To mitigate some of these drawbacks, urban mobility in many cities has equipped itself to create special spaces where scooters can circulate. However, these interventions have proved to be unsuccessful, especially from the point of view of safety, in order to counteract the danger to pedestrians and the riders themselves.

There is therefore a need to define an innovative electric scooter equipped with devices for the safety and reliability of the scooter itself that would eliminate or at least minimize the aforementioned drawbacks.

Summary of the Invention

In order to substantially solve the technical problems outlined above and, in particular, to increase safety in the use of electric scooters, an object of the present invention is to define an electric scooter provided with an anti-collision system and related control strategy thereof.

A further aim is to adopt artificial intelligence in micro-mobility vehicles. The challenge is to use standard devices with low cost and performance in order to provide the driver with advanced functionalities that are typical of large vehicles but, on such vehicles, can only be integrated at high cost. However, it is not intended as 'safety critical' functionality and does not interact in any way with driving.

The aim of the control strategy of the anti-collision system is to optimize the processing of the acquired images and the activation of the anti-collision system. According to further purposes of the present invention, the scooter is provided with further systems which confer additional advantages in terms of safety and/or reliability. In particular, the scooter is provided with a removable battery which can be easily replaced and/or recharged separately from the scooter. Thus, the present invention relates to an electric scooter comprising an anti-collision system, having the characteristics set forth in the independent product claim appended hereto.

Further, the present invention relates to a method for operating the anti-collision system. Further preferred and/or particularly advantageous embodiments of the invention are described according to the features set forth in the appended dependent claims.

Brief Description of the Drawings

The invention will now be described with reference to the appended drawings, which illustrate some non-limiting examples of implementation, wherein:

- figure 1 is a side view of an electric scooter provided with the anti collision system according to the present invention,

- Figure 2 is a rear view of the electric scooter of Figure 1, - Figure 3 is a side view of the electric scooter of figure 1, in a closed configuration,

- figure 4 is a rear view of the electric scooter of figure 1, in a steering configuration,

- figure 5 schematically illustrates a detail of the electric scooter handlebar of figure 1, and in particular the housing containing the battery,

- figures 6 and 7 schematically illustrate a detail of the battery extraction device referred to in figure 5, in open and closed configuration respectively,

- figure 8 schematically illustrates the hardware of the anti-collision system present on the electric scooter;

- Figure 9 is a schematic of the high level architecture of the anti collision system software of Figure 8;

- Figure 10 is a block diagram of the method of managing and analyzing a video image according to the present invention, - Figure 11 is a block diagram of the method for managing collisions on a 2D image according to the present invention,

- figure 12 is a logic diagram of the method for managing artificial intelligence models according to the present invention.

Detailed Description By way of non-limiting example only, the present invention will now be described by reference to the above-mentioned figures.

In particular, Figures 1-7 illustrate an electric scooter 100 according to the invention. The electric scooter 100 comprises a platform 1, on which the feet rest, a handlebar 2, pivoted on the platform 1, which is used to change direction, at least one brake 3, disc and/or electric, located on the front wheel and three wheels 4, 5, 6, one front wheel 4 and two rear wheels 5, 6. All wheels 4, 5, 6 are puncture-proof, through the use of solid honeycomb tyres at the rear or 'foam' inserts instead of an inner tube at the front. As can be seen in Figure 4, the rear axle 60 is composed of a truck (equivalent to those used in skateboards) that allows the rear wheels 4, 5 to steer, overcoming the limitation of current scooters related to turning radius.

Advantageously, the frame of the electric scooter 100 is made of aluminum and platform 1 is made of wood. The width of the platform 1 is optimized to allow the scooter to be ridden even with parallel feet.

Handlebar 2 includes an aluminum front tube 2' containing a removable battery 7. On the handlebar 2, a display or smartphone 8 is centrally positioned: it indicates speed, riding mode, and battery charge status. The battery 7 can be removed from the front tube 2' and is available in different 'sizes', depending on the desired driving range. We are therefore talking about the 'modular' battery concept, whereby batteries of different sizes share the same external 'case' so as to be completely interchangeable.

The extractable battery, again with a view to safety and reliability, makes it possible to extend the vehicle's driving autonomy, when, for example, a spare battery is carried along that can be easily replaced in place of an exhausted one. If you do not wish to carry a second spare battery on board, it will still be easy to obtain a replacement battery at a charging station or service station, or to quickly recharge the existing battery. As shown in figure 3, the electric scooter 100 is foldable thanks to a hinge 10 located near the lower portion of the front tube 2', above the front wheel 4, which allows the closure of the tube 2' itself, above the platform 1. Once closed, the scooter 100 can be transported like a trolley thanks to the two rear wheels resting on the ground. Advantageously, the electric scooter 100 remains upright on its own, thanks to the weight distribution (centre of gravity) and the geometry of the rear part, which then becomes a base. As well as providing undoubted convenience in use, these latter features also increase the safety of the scooter by preventing accidental falls and consequent breakdowns or, even worse, injuries.

The motor of the scooter is of the "in-wheel motor" or "hub-motor type", i.e. it is located in the hub of the front wheel, it is an off-the-shelf electric motor. Brake 3 is of the front disc (safety) type with an electric motor brake. When braking, the motor is switched off for safety. The brake lever and throttle lever are located on the handlebar.

The electric scooter 100 also includes rear position indicators that can be operated by the switches on the handlebar.

As shown in figure 5, a removable battery 7 positioned inside the front tube 2' is closed by a flap 9 on which the display or smartphone 8 is positioned. As shown in figures 6 and 7, the battery 7 is removed and re inserted by means of a rotation of a cap 11 equipped with a spring, integral with the battery 7. The scooter 100 also has a front LED light and a rear LED strip for the stop and brake light and indicator lights. From a safety perspective, there is no doubt that the greatest advantage of the electric scooter according to the present invention is that it is equipped with an anti-collision system comprising a virtual rear-view mirror and a system for generating anti-collision alarms.

As shown in Figure 8, the virtual rear-view mirror comprises a camera 12 positioned at the rear of the scooter 100, constrained to the footboard. The camera hardware generates its own WIFI network in 'access point' mode. A smartphone 8 (WiFi Client), via a mobile application, connects to the camera 12 WiFi network. The connection is standard and takes place via network identification via SSID and the use of a possible password. The control strategy is developed according to an optimal and discreet selection of images recorded by the virtual rear-view mirror, i.e. a camera positioned at the rear of the scooter and connected via WI-FI with a screen or smartphone of the user. In particular, the camera is able to send images that are processed by the system and, depending on the need, activate an anti-collision warning system.

The mobile application for smartphones makes it possible to

- change the scooter's settings (maximum speed and driving mode),

- activate and deactivate the virtual rear-view mirror mode and the anti-collision warning system, as well as display the image of the rear road and alerts (both visual and audio) on your screen

- have a navigation system integrated into the application,

-other functions, such as the purchase of spare parts, etc.

Having established the WIFI connection (Client and Server), the mobile application takes care of establishing a communication channel via standard protocols (UDP) and initializing the Mobile APP-Camera communication. A data 'stream' is established. The mobile application acquires video frames from the ’stream’ and shows a preview of them on the display of the smartphone 8, simulating a virtual rear-view mirror. The camera 12 offers a 120° wide view in the horizontal plane to give the driver a greater field of vision.

Advantageously, the mobile application is able to run artificial intelligence models using ’Tensor Flow Lite by Google’. Specifically, Object Detection’ models are used.

The anti-collision alarm generation system is the result of this processing and leads to the generation of three alarms (visual/sound) for the driver, for example:

1. no danger -> no indicator or sound,

2. object approaching from the rights danger indicator "Right" (first alarm signal), 3. object approaching from the left^ "Left" danger indicator (second alarm signal),

4. object approaching from the centre/undetermined situation -> ’Danger’ indicator (third warning signal).

As the block diagram of the high-level software architecture in figure 9 illustrates, this consists of the following modules:

- S: ’Streaming’ module, which is responsible for establishing the Wi Fi connection between the mobile application and the hardware; from this module, images are extracted at the ’frame’ level, according to a specified ’frame rate’ (e.g. 20 FPS); - AI: Artificial Intelligence Model Execution module, in which the extracted images become input for the artificial intelligence engine; the execution of an 'object detection' model generates a collection of identified objects on each image. Processing takes place in the background at a time independent of the 'frame rate';

- the outputs of the artificial intelligence model for each image become inputs for analyzing the CDP ('Collision Detection and Processing') position of objects on the scene, understanding possible dangerous situations and handling alarms;

- a graphical UI and sound part allows the user to be alerted to possible dangerous situations via the mobile application.

The method of management of the anti-collision system is illustrated in the flow chart in Figure 10. Each image is processed according to the following steps: a. connecting S10 the camera and start streaming; b. acquiring S20 a new image from the streaming; c. evaluating S30 whether the AI module is engaged: if no, return to step b., if yes, continue; d. running S40 an "object detection" model; e. checking S50 for possible collision with a grid; f. carrying out S60 an analysis of the potential risk of collision; g. evaluating S70 if there is a risk of potential collision: if no, return to point b., if yes, continue; h. generating an alarm signal.

Regarding point c., the choice depends on the following considerations: when there is a new frame (the period between frames depends on the 'framerate'), it is processed if and only if the processing and 'post processing' of the AI module of the previous frame have been completed. Otherwise, the new frame is not processed. As soon as the processing is complete, the most recent and newly received frame is processed.

Thus, the slower the processing and computing power of the device, the more frames will be discarded. At best performance will result in AI module processing and 'post-processing' every l/'framerate' seconds. In any case, the video display is never blocked by processing, and remains fluid, preserving the virtual mirror functionality.

• The method aims to:

• - avoid performance-intensive and complex 'tracking' on a system with hardware that does not perform well in terms of 'frame rate', resolution and image quality, · - create customized profiles of areas "sensitive" to the presence of objects on the scene,

• - create a generic software module on which to build subsequent risk analysis and alarm generation policies.

• The purpose of the configuration is to virtually draw a "Box Collider" mask on the scene depending on the type of vehicle and the type of analysis to be performed. The mask is drawn on each image: varying the mask varies the portions of the image to be analyzed. There are several possible configurations depending on:

- group: 'boxes' may belong to groups. A group, for example, could be a set of Collider boxes placed in a row on the image, - height of the "boxes": the height of the boxes allows the proportions and perspective of the 2D image to be managed. Ideally, the boxes should cover the ground in perspective, assuming the camera is parallel to the ground, - the number of "boxes": the number of boxes generates a finite discrete set of possible collisions on the scene. Increasing the number of boxes results in a denser set of areas over which collisions can be identified.

In particular, by "Box Collider" we mean a rectangle on the scene that occupies a part of it; by "Bounding Box" a rectangle around an object identified by the Artificial Intelligence; by "Box Collider Group" a set of "Box Colliders" that identify a given area on the image and by "Collision" a "Software" event that is generated following an intersection between a "Box Collider" and a "Bounding Box". Each 'Box Collider' group can identify a collision with a 'Bouding Box'.

In particular, a collision results in:

- a collision if a "Bouding Box" intercepts a "Box Collider" in the drawn mask;

- the collision provides the group(s) involved in the collision (e.g. Row 1, Row 2, Row 3);

- the collision provides the identifying "id" of each Box intersected by the "Bouding Box" in the collision (e.g. Box 1 to Box 4). According to the present invention, a method for image management is illustrated in Figure 11, comprising the following steps: i. viewing S100 a new image, ii. drawing S110 a collision mask (Collision Matrix), iii. detecting S120 a rectangle around an object ('Bounding Box'), iv. evaluating S130 the collision between a 'Colliders Group' and a

'Bounding Box', v. if the possible collision is evaluated S140 negatively, go back to step iii., if it is evaluated S140 positively continue, vi. saving S150 all data, vii. running S160 an algorithm, viii. collecting S170 data. The results of steps i. - v. are used as input for algorithms in step g., which process the results.

Advantageously, each algorithm can exploit the collision information of both a single image and, operating over time, several consecutively processed images. Advantageously, by storing collision data of areas and objects in the received image stream, various types of analysis can be implemented, exploiting the plasticity of this method.

Collision history management makes it possible to:

- keep the Box IDs in memory and carry out discrete tracking by replacing the "pixel" with the "Box", only in the areas of interest in the framed image and not in the whole,

- carry out a sort of tracking of the areas of interest "affected" by the objects. The greater the number of 'Box Colliders' drawn, the greater the granularity of the tracking, - understand if objects are approaching by analyzing the areas closest to the vehicle,

- understand if objects are receding by colliding with more distant areas,

- exclude numbers and/or groups of Box Colliders easily (e.g. areas too far away or not of interest),

- shift algorithm management to areas and not to identified objects.

This avoids problems with loss of object identification especially when using low-cost cameras (low quality, resolution and frame rate).

As illustrated in the logic diagram in Figure 12, once the objects are identified on the images, the mobile application processes them according to their position relative to the d rive r/ve hide.

Image processing by means of artificial intelligence models is carried out by means of the following steps:

- acquiring S200 images by means of artificial intelligence models, - generating S210 an output signal as a list of detected objects,

- applying S220 a filter only on the desired objects,

- detecting S230 a possible collision,

- generating S240 an output signal as a list of data on objects in possible collision (distance, coordinates, etc.). In addition to the modes of implementation of the invention, as described above, it is to be understood that there are numerous further variants. It should also be understood that said modes of implementation are merely illustrative and do not limit the scope of the invention, nor its possible applications or configurations. On the contrary, although the above description makes it possible for a person skilled in the art to implement the present invention at least according to an illustrative configuration thereof, it should be understood that numerous variations of the described components are conceivable, without thereby departing from the scope of the invention as defined in the appended claims, interpreted literally and/or according to their legal equivalents.