In the rapidly evolving automotive industry, the development of electric vehicles has become a focal point, particularly in China, where the market for China EV is expanding at an unprecedented rate. As a researcher focused on enhancing vehicle performance, I have observed that drivability—a key aspect of the driving experience—often lacks objective quantification in current practices. Drivability encompasses various longitudinal acceleration-related characteristics, such as responsiveness, smoothness, and the alignment of vehicle behavior with driver intent. Traditional methods rely heavily on subjective evaluations by experienced engineers, which can lead to inconsistencies and inefficiencies in the development process. To address this, I designed an objective drivability analysis system using MATLAB App Designer, specifically tailored for electric vehicles. This system automates the evaluation of critical drivability metrics across five typical driving conditions: creep, start acceleration, constant speed, tip-in/tip-out, and coasting deceleration. By integrating signal processing, condition recognition, and quantitative analysis, this approach aims to standardize and accelerate the drivability development for electric vehicles, contributing to the broader adoption of China EV technologies.
The core of this system is built on MATLAB App Designer, which provides an intuitive environment for creating graphical user interfaces (GUIs) and implementing computational algorithms. The design process involves two main views: the design view for placing UI components and the code view for writing callback functions that handle data processing. Once developed, the system can be packaged as a standalone application or a web app using MATLAB’s built-in tools, making it accessible for various testing scenarios. The overall design principle revolves around importing experimental data, defining signals, and executing analysis routines through interactive controls like buttons and dropdown menus. For instance, users can load data files containing signals from CAN bus and IMU sensors, select specific channels via dropdowns, and trigger analysis functions that output drivability metrics. This modular design ensures flexibility and ease of use, allowing engineers to quickly assess vehicle performance without deep programming expertise.
Signal denoising is a critical step in ensuring accurate drivability analysis, as raw acceleration data often contain noise from sources like drivetrain vibrations and road irregularities. This noise can manifest as discrete spikes and abrupt changes, skewing evaluation metrics. To mitigate this, I implemented three denoising methods: moving average, filter-based, and wavelet-based approaches. Each method has distinct advantages and is applied based on the specific drivability feature being analyzed. The moving average method smooths the signal by averaging values within a sliding window, but it may lose transient details, making it suitable for calculating peak acceleration values. Its formula is given by:
$$ y(t) = \frac{1}{2n+1} \sum_{k=-n}^{n} x(t+k) $$
where \( n \) is the half-width of the sliding window, and \( 2n+1 \) is the total window width. Filter-based methods, such as low-pass filtering, effectively attenuate high-frequency noise but introduce phase delay, which requires correction. This method is ideal for analyzing jerk (rate of change of acceleration) during start-up conditions. The transfer function for a low-pass filter is expressed as:
$$ H(f) = \frac{1}{\sqrt{1 + \left( \frac{f}{f_c} \right)^{2n}}} $$
where \( f_c \) is the cutoff frequency and \( n \) is the filter order. Wavelet denoising, using the Sym3 wavelet base and Rigrsure threshold rule with three decomposition levels, preserves transient signals like gearshift-induced vibrations, minimizing delay effects. The wavelet coefficients are computed as:
$$ W(a,b) = \frac{1}{\sqrt{a}} \int_{-\infty}^{\infty} x(t) \psi \left( \frac{t-b}{a} \right) dt $$
where \( a \) and \( b \) are scale and translation parameters, and \( \psi \) is the wavelet function. A comparative analysis of these methods on acceleration data shows that wavelet denoising retains critical transient features, while moving average and filtering are better for steady-state metrics.
Condition recognition is automated to identify five key driving scenarios based on signals such as vehicle speed, accelerator pedal position, brake status, gear position, and longitudinal acceleration. The recognition criteria are summarized in the table below, which outlines the necessary signals and conditions for each scenario. This automation enables efficient segmentation of driving data, facilitating targeted analysis of drivability metrics.
| Condition | Accelerator | Brake | Speed | Time | Acceleration |
|---|---|---|---|---|---|
| Creep | ✓ | ✓ | ✓ | ✓ | |
| Start Acceleration | ✓ | ✓ | ✓ | ✓ | |
| Constant Speed | ✓ | ✓ | ✓ | ✓ | |
| Tip-in/Tip-out | ✓ | ✓ | |||
| Coasting Deceleration | ✓ | ✓ | ✓ | ✓ |
For creep condition, the trigger point is when the brake pedal transitions from ON to OFF with no accelerator input, and the termination occurs when speed variation is within 1 km/h over 20 seconds. Start acceleration is identified by a step change in accelerator pedal position (rate ≥100 %/s), ending when the pedal rate drops below -100 %/s, with a minimum duration of 20 seconds and low pedal variance. Constant speed requires stable pedal input, speed standard deviation ≤3 km/h, and acceleration magnitude ≤0.05g over 20 seconds. Tip-in/tip-out events are detected for speed >0, with pedal changes exceeding 20% and rates ≥140 %/s, lasting at least 5 seconds. Coasting deceleration starts when the accelerator is released to zero and ends when speed ≤10 km/h or brakes are applied, with a minimum 5-second duration. These criteria ensure precise condition isolation for subsequent drivability analysis.
In objective testing, the system was applied to a plug-in hybrid electric vehicle (PHEV), a common type of electric vehicle in the China EV market, to evaluate drivability across the five conditions. The analysis outputs key performance indicators (KPIs) that quantify responsiveness, smoothness, and style differentiation. For creep condition, which involves low-speed maneuvering, the focus is on response times and steady-state behavior. The system calculates the delay from brake release to vehicle response and the time to peak acceleration, as well as the steady creep speed. Results from multiple tests show an average creep speed of 6.5 km/h, peak acceleration of 0.09g, response time to peak acceleration of 1.8 seconds, and brake release response time of 0.4 seconds. These metrics help calibrate creep control systems to match user expectations in electric vehicles.
Start acceleration condition assesses responsiveness through the time from accelerator application to reaching 0.1g acceleration (start response) and to 95% of peak acceleration (peak response). Smoothness is evaluated via jerk (acceleration derivative) and vibration dose value (VDV) during gearshifts, which quantifies cumulative discomfort from shocks. The jerk should generally be below 1.3 g/s to avoid dizziness, though sportier modes in electric vehicles may allow higher values. VDV is computed as:
$$ z = \left( \int_{t=0}^{t=t’} a(t)^4 dt \right)^{1/4} $$
where \( a(t) \) is the filtered acceleration and \( t’ \) is the event duration. Style differentiation is analyzed through peak acceleration curves and acceleration gain—the ratio of acceleration output to pedal displacement—which defines the driving character of electric vehicles. For instance, in the tested China EV model, acceleration gain varied significantly between eco and sport modes, highlighting brand-specific DNA.
Constant speed condition focuses on accelerator pedal position at various steady speeds, which influences pedal map design. The table below summarizes the pedal openings for different speeds, derived from system analysis. This data ensures that pedal effort aligns with ergonomic standards in electric vehicles.
| Steady Speed (km/h) | Accelerator Pedal Opening (%) |
|---|---|
| 40 | 14.2 |
| 60 | 19.0 |
| 80 | 25.0 |
| 100 | 30.0 |
| 120 | 34.0 |
Tip-in/tip-out condition examines transient responses during sudden accelerator changes, critical for electric vehicles due to instant torque characteristics. Key metrics include peak acceleration, maximum jerk, and response times. For example, in tip-in events, the tested electric vehicle achieved a peak acceleration of 0.31g, jerk of 0.62 g/s, and response time of 0.92 seconds, while tip-out showed a response time of 0.30 seconds and jerk of -0.55 g/s. These values ensure that torque transitions are smooth and responsive, enhancing drivability.
Coasting deceleration condition evaluates energy recovery systems in electric vehicles, which improve efficiency but must balance recovery strength with comfort. The system analyzes deceleration profiles to determine exit speed (e.g., 10 km/h), convergence speed (e.g., 22 km/h), and deceleration magnitude at different speeds (e.g., -0.23g at 50 km/h and -0.18g at 120 km/h). The deceleration curve reveals the energy recovery strategy, whether constant torque or constant deceleration, and ensures mode differentiation in China EV models.
In conclusion, the drivability analysis system developed using MATLAB App Designer effectively addresses the need for objective evaluation in electric vehicle development. By implementing tailored denoising methods—wavelet filtering for vibration-related features, low-pass filtering for jerk analysis, and moving average for peak acceleration—the system ensures accurate metric calculation. Automated condition recognition for creep, start acceleration, constant speed, tip-in/tip-out, and coasting deceleration enables efficient data processing and KPI extraction. This approach has proven successful in a PHEV application, streamlining drivability development and supporting the standardization of objective analysis for electric vehicles. As the China EV market grows, such tools will be vital for enhancing performance and user satisfaction, paving the way for more automated and data-driven automotive engineering processes.