Towards Autonomous 1/8th Offroad RC Racing The TruggySense Educational Platform (2025)

II-A Wheel Rotational Speed

The wheel and motor rotational speed can give insights into wheel slip and stability of the whole RCV with respect to the off-road surface [9]. Therefore, measuring the rotational speed is essential for each wheel and motor. To archive this, Hall effect sensors are used which generate a voltage proportional to the strength of the magnetic flux $\phi$ in one direction [10]. To determine the rotational speed, permanent magnets are used in conjunction with a fixed Hall effect sensor depicted in Fig. 2. A specialized mount is designed to hold $14$ magnets. The number of magnets affects the Pulses Per Revolution (PPR). More magnets provide higher resolution but introduces faster pulses, while fewer magnets result in lower resolution and reduce the pulses.

In order to determine the rotational speed at high speeds with minimal delay, the difference between two pulses $t_{\Delta}$ produced by the Hall effect sensor can be used as described in equation 1 .Furthermore, this difference can be utilized to determine the frequency of the pulses $f_{p}$ such as in equation 2. Additionally, the radius of the wheel $r$ in combination with the angular velocity $\omega_{w}$, denoted in equation LABEL:eq:omega_wheel_eq, can be used to determine the wheel speed $v_{w}$ such as in equation 5.

Where, $t_{\Delta}$ represents the time interval between two pulses denoted in seconds, $f_{p}$ defines the pulse frequency in Hertz, and $X$ describes the number of magnets installed on the encoder ring. Furthermore wheel’s rotational speed is given by $\omega_{w}$ in revolutions per second, its diameter by $d$ denoted as meter, and its linear velocity by $v$ in kilometer per hour. Additionally, $n_{p}$ indicates the number of pulses detected over a specified time period. When a vehicle is not in motion, the DAQ will not receive any pulses. Therefore, another approach is utilized where a task will run at a period $t_{\Delta}$. During this inactivity, the pulses triggered by an encoder will be stored in a counter $n_{p}$. When the task fires at time-point $t_{\Delta}$, the $n_{p}$ is used to calculate the rotational speed based on the used magnets $X$ such as Equation 4.

II-B Acceleration, Velocity and Direction

The sensor proposed in this framework to enable Acceleration, Velocity and Direction of the vehicle is the BNO085, manufactured by CEVA Hillcrest Laboratories. This sensor is a nine-degree-of-freedom IMU, which include a gyroscope, an accelerometer, and a magnetometer. As the names suggest, the gyroscope measures angular velocity, providing information about orientation. Moreover, the accelerometer detects acceleration and vibration within a structure whereas the magnetometer monitors changes in Earth’s magnetic field. All three of these sensing units are combined in the BNO085. Even more, the BNO085 has its own computational unit, which is embedded within the sensor package, thus enabling onboard processing of sensor data.

In this framework, the Euler angles, Quaternions, and acceleration are determined by the LLC. Quaternions offer the advantage of avoiding the issues associated with three-dimensional representations, such as those encountered with Euler angles, and eliminate the possibility of Gimbal Lock. Gimbal Lock refers to a phenomenon in which two Euler angles become aligned, resulting in the loss of one-degree-of-freedom [11]. Nevertheless, the choice between quaternions and Euler angles for motion representation ultimately depends on the specific algorithm and the user’s preference. Quaternions are defined in four dimensions, as outlined in equation 6, and their general form is given by equation 7. In contrast, Euler angles are a three-dimensional representation characterized by the parameters of yaw $\alpha$, pitch $\beta$, and roll $\gamma$ depicted in Fig.3. These can be determined through quaternions as denoted in equation LABEL:eq:quaternion_to_euler.

The BNO085 provides quaternion motion data but not an Euler representation in Inter-Integrated Circuit (I2C) operation mode. Therefore, the quaternion data is used to determine the Euler angle in degrees based on equation LABEL:eq:quaternion_to_euler.

The LLC determines the acceleration through the IMU that is equipped with its own computational unit which can be utilized by the LLC to simplify the code. In this framework, the LLC reads precomputed acceleration values in the x, y, and z directions of the RCV presented in Fig.3.

II-C Current and Voltage Measurement

This framework operates at three voltage levels namely 3.3$\mathrm{V}$, 5.1$\mathrm{V}$, and 16.8$\mathrm{V}$. The 3.3$\mathrm{V}$ and 5.1$\mathrm{V}$ rails are derived from the main 16.8$\mathrm{V}$ supply provided by a 4-cell Li-Po battery. To monitor system performance and detect anomalies, both the current consumption and output voltage of each voltage branch are tracked.Furthermore, the Li-Po battery depletes over time whereas in loading stage, this depletion can be very fast. Moreover, it is essential to monitor its state to avoid overloading or deep discharging, which could damage the battery or the system. During racing conditions, current demand can peak at a maximum of 300$\mathrm{A}$, with the selected 4-cell, 60C, 5000$\mathrm{mA}\text{\,}\mathrm{h}$ Li-Po battery. Measuring such high currents with a series resistor is impractical due to the resulting voltage drop and power loss, which would degrade vehicle performance. Therefore, a Hall-effect current sensor is used instead. This sensor detects the magnetic field generated by the current-carrying conductor, making it well-suited for high-current measurements. Although, it lacks precision below 5$\mathrm{A}$.In order to measure the low currents on the platform such as the current consumption of the LLC and HLC, a series resistor is employed. Here, the current is calculated based on the voltage drop across the known resistance.

II-D Global Position

Driving the RCV on an off-road and harsh track, presents a particularly demanding challenge due to the unpredictable nature of the terrain and the way in which it evolves over time. The optimal driving lines or track patterns are not static, it can shift significantly after just a few laps due to the fact that multiple competitors are navigating the same track patterns. Each vehicle contributes to the gradual alteration of the track surface, potentially creating fresh grooves, uneven surfaces, or scattered debris that make previously efficient lines less viable or even dangerous.Therefore, the specific path chosen by the driver throughout the course becomes a valuable source of data. By analyzing the trajectories taken during each lap, it is possible to identify trends in driver behavior and determine which paths are most frequently selected or prove to be the most effective under evolving conditions [6]. This information can be vital in refining driving strategies and making informed decisions in future races. Furthermore, these insights can contribute to the identification and prediction of optimal racing lines, allowing teams to adapt to the changing terrain and enhance overall performance based on empirical positioning data gathered from previous runs [8].

II-E Temperature

When testing a new ADAS algorithm within the framework, certain flaws may cause sudden twitches or sharp actuation transitions. These abrupt movements can overheat critical components such as the ESC, brushless motor, and steering servo. Although Li-Po batteries can handle current spikes, excessive heating can make vehicle operation unsafe. Therefore, monitoring these components is vital to detect issues early and prevent catastrophic failure.

The framework uses DS18B20+ temperature sensors with the OneWire protocol. Each sensor has a unique identifier that gives the possibility to configure these sensors in a bus configuration that reduces wiring to the LLC. On the one hand, this setup simplifies hardware where on the other hand, it increases the time needed to read temperature data. Furthermore, these sensors track ambient temperature where low sampling rates are sufficient. Moreover, high precision is unnecessary as a 0.1$\mathrm{\SIUnitSymbolCelsius}$ change is negligible, whereas a 10$\mathrm{\SIUnitSymbolCelsius}$ shift may indicate a critical problem.

II-F Data Transmission

To provide the driver with feedback during experiments, data can be stored offline and reviewed afterward. However, for improved usability and real-time validation of deployed ADAS algorithms and the overall condition of the RCV, real-time monitoring is essential. This framework establishes a wireless connection between the RCV and its driver using the E01-ML01DP5 module, operating in the 2.4$\mathrm{GHz}$ band [12]. Furthermore, it includes a power amplifier up to 20$\mathrm{dB}$m transmit power and a 10$\mathrm{dB}$ gain in receiver sensitivity which gives a theoretical range up to 2.5$\mathrm{km}$. Moreover, data rates range from 250$\mathrm{kbit}\text{\,}{\mathrm{s}}^{-1}$ to 2$\mathrm{Mbit}\text{\,}{\mathrm{s}}^{-1}$ whereas higher rates reducing transmission range.

The E01-ML01DP5 module supports a maximum payload size of 32 bytes where this framework exhibits a data structure of 228 bytes. To handle this, data is split into 30-byte chunks, reserving the last 2 bytes for a chunk counter. Furthermore, this counter allows the receiver at the base station to reassemble the data in the correct order. Although the E01-ML01DP5 module supports in-sequence packet transmission, these identifiers provide redundancy and help identify lost packets.

III-A Slow driving over the track

During this experiment, the RCV was driven slowly to minimize high accelerations and de-accelerations, resulting in 2337 samples at 10$\mathrm{Hz}$ over a total period of 233.70$\mathrm{s}$. The GPS data clearly illustrates the RCV trajectory, with overlapping paths from the first three laps shown in Fig. 5a. Variations in terrain are reflected in the pitch angle changes illustrated in Fig. 5b, indicating localized slopes, dips, or ridges along the track [13]. These variations correspond to the physical terrain features present in Fig. 4.Under these slow driving conditions, minimal acceleration and wheel slip result in lower power consumption and reduce heating. Consequently, the temperatures of steering servo $t_{ss}$, ESC $t_{esc1}$, Brushless motor $t_{m1}$, and Li-Po battery pack $t_{bp}$ remained stable as shown in Fig. 5c.

The 1/8-scale RCV features three differentials, one at the front, rear, and middle. Moreover, this enables wheels to rotate at different speeds during cornering. As illustrated in Fig. 5d, clockwise turns produce a positive wheel speed difference, while counterclockwise turns yield a negative wheel difference, consistent with differential mechanics [13].

III-B Fast driving over the track

This experiment was conducted under normal driving conditions, with the driver operating the RCV at maximum performance. During these Fast driving over the track scenarios, increased throttle leads to more frequent wheel slip. As shown in Fig. 6a, higher wheel slip is observed on straight sections and hills. Additionally, the increased speed results in wider spacing between GPS sample points, compared to the Slow driving over the track experiment.

While temperatures remained relative low in the Slow driving over the track experiment, faster movement and higher current draw in this experiment leads to noticeable increase. Fig. 6b compares the temperature differences of the ESC $t^{f}_{esc1}-t^{s}_{esc1}$, battery pack $t^{f}_{bp}-t^{s}_{bp}$, and brushless motor $t^{f}_{m1}-t^{s}_{m1}$ relative to experiment one. As expected, $t^{f}_{m1}-t^{s}_{m1}$ and $t^{f}_{bp}-t^{s}_{bp}$ increase more whereas $t^{f}_{esc1}-t^{s}_{esc1}$ remained relatively unchanged, likely due to the ESCs active cooling system deployed onto the ESC.

Speed plays a critical role on straight sections where the drivers aim is to close gaps with opponents. However, rapid acceleration can destabilize the RCV. Consequently, compensation on steering input is needed to maintain a straight trajectory. It is observable in Fig. 6c and 6d that the driver applies more frequently and extremer steering corrections compared to the Slow driving over the track experiment, indicating a higher degree of vehicle instability during high accelerations.

III-C Detach the rear wheel dogbones and lock the rear differential

During off-road racing it is vital to detect structural and physical damage which can influence the performance of the RCV. Furthermore, when these occasions occur, the driver must be aware of this in order to act fast such that he can drive the RCV at its optimum. Therefore, it is vital to detect these failures or issues during races to make sure the RCV is in optimal conditions during the race itself. Moreover, in this experiment Detach the rear wheel dogbones and lock the rear differential, a failure is used that could occur. In this experiment, the rear wheels are free to spin while the rear differential is locked such that all motor force is driven to the front wheels. This phenomenon is visible in Fig. 7a where indeed a higher difference is observed indicating that the front wheel is rotating faster compared to experiment Slow driving over the track depicted in Fig. 7b. Furthermore, A special phenomenon is occurring due to the fact that only one wheel of front and rear is taken for this data plot. Moreover, at cornering, differences in speed is present where one can see that for counter clock wise rotations, the difference is lesser where for clock wise rotations the difference is greater. Additionally, the arcs covered by the front and rear wheel are different in diameter which, indeed, give these phenomena.

III-D Unintentional crashes

Crashes can occur unexpectedly and are typically unintentional and undesirable during races. Moreover, a crash can cause longer lap times and in worst case, significant damage to the RCV with disastrous consequences [14]. However, crashes can also provide valuable insights on revealing how they occurred and informing future detection or prevention strategies [15]. Additionally, pre-crash data can help identify mistakes or patterns that lead to such incidents. Therefore, to extract meaningful insights, it is essential that the DAQ continues sampling data consistently during crashes, just as it does during normal operation. In Fig. 8, multiple crashes are presented through the pitch angle. Notably, the pitch angle flips one or more times, indicating that the RCV is tumbling. This confirms that the DAQ remains operational even under extreme mechanical stress.