A human-centered HMI concept that restructures control access, task depth, and system feedback to reduce complexity across the in-car experience.
The modern vehicle is shifting from a mechanically defined product to a connected, continuously updated digital system. As intelligent features, real-time status layers, and cross-module services continue to grow, the in-car interface takes on a more direct role in how drivers access information, complete tasks, and maintain attention in motion.
This project explores how an EV cockpit can be reorganised around clearer control access, shallower task paths, and more legible system feedback. Through user research, structural design, and high-fidelity prototyping, it proposes a human-centered HMI system that reduces complexity while supporting safer, more intuitive, and more coherent interaction across the in-car experience.
As software scale, feature density, and update frequency continue to grow, vehicles are evolving from mechanical products into continuously updated digital systems. HMI is therefore being pushed into a more central role, carrying not only functional complexity, but also user experience and brand perception.
Vehicles are no longer mechanical products; they are evolving into mobile digital ecosystems.
As software continues to enter the vehicle system, more complexity is being brought directly into the driving process. As capabilities expand, the interface is expected to support information transfer, state confirmation, and task switching within a much shorter time window, causing complexity, efficiency, and safety to collide within the same layer of experience.
To give the following research and design stages a clear problem boundary, the project examines this set of tensions through four dimensions: cognition, interaction, human factors, and driving context. Together, they determine which tasks need shallower paths, which states need clearer explanation, and which information should reorganise itself as the context changes.
Cognitive Foundations
Understand human cognition
Human attention has strict limits; multitasking can delay reaction time by ~0.7–1.3s.
Multiple cognitive resources (visual, auditory, cognitive) enable multimodal interaction to distribute workload.
Situation awareness develops across perception, comprehension and projection stages.
Design implication
Interfaces should prioritise critical information and support rapid visual scanning with minimal cognitive load.
Research basis: Kahneman (1973) • Wickens (2008) • Endsley (1995) • Norman (2013)
These questions extend the tensions identified in background research, but at the user level they become more specific pressures around efficiency, safety, discoverability, and structural continuity.
These sources did not lead to a single direct answer. Instead, they were consolidated into representative user models, recurring scenario paths, and a clearer set of design judgements.
A single HMI does not serve a single user type. It has to accommodate different expectations around efficiency, safety, and discoverability at the same time.
Family travel
lower tech acceptance
values reassurance
Lower technology acceptance and a stronger need for reassurance and discoverability.
Main need
Core functions should be immediately understandable, with fewer repetitive adjustments.
Design implication
Complex interfaces can directly lead to avoidance. Core functions need to be easier to find, and repeated setup should be reduced.
These representative models do not describe aesthetic preference. They show that one HMI system must support three different kinds of demand at once: efficiency, safety, and reassurance.
In automotive contexts, task priority shifts with driving state. More important than functional categorisation is how driving, parked, and transitional states reshape attention and information density.
Scenario 02
Task state
Once driver assistance intervenes, the system requires the driver to rebuild situational awareness and complete a takeover.
Key obstacle
The issue is not only that an alert appears, but whether the driver can quickly understand the current state and next action.
Design implication
Safety-critical tasks need to prioritise clarity of explanation, with more direct state and feedback hierarchy.
Scenario Findings
The real issue is not a lack of functions, but the fact that short-form tasks are buried too deep.
The real risk comes from states that cannot be interpreted quickly, not from taking one extra tap.
From parked to driving, and from module to module, continuity matters more than interface richness.
At this stage, research is used to define judgement boundaries first, leaving concrete interface decisions to the following structure and high-fidelity stages.
Design Judgement
Parked, driving, and module-switching states need to reorganise information within one shared structural framework, reducing mode-switching cost.
This stage ultimately clarified three things: what should become shallower, what should become clearer, and what should reorganise itself with changing context. In the next stage, those judgements continue into information architecture, user flow, wireframes, and high-fidelity interface design.
At this stage, research is used to define judgement boundaries first, leaving concrete interface decisions to the following structure and high-fidelity stages.
In the driving context, deeply nested menus are dangerous. For this HMI system, I adopted a shallow hierarchy strategy:
High-Frequency Actions: Placed in the 'Control Centre' for 1-tap access (e.g., Quick Toggles, Volume).
Configuration: Detailed setups (e.g., ADAS, full Lighting customization) are housed in 'Settings', accessible only when parked or safe. This separation ensures that 80% of tasks are achievable within 2 taps, minimizing driver distraction.
After defining the Information Architecture, I used flow mapping to test whether tasks were actually placed at the right level. The goal was not to make everything one tap away. It was to match path depth to task frequency, visibility needs, and safety risk.
Once task paths have been distributed across different levels, the page skeleton needs to carry those decisions forward. At this stage, wireframes and layout strategy are used to stabilise entry depth, state transitions, and module expansion within one shared page framework, and to test whether that framework can support different driving contexts and task types.
The centre console display is the nerve centre of the entire HMI, serving as the primary gateway for users to quickly access all functions. The interface intelligently adapts by automatically switching between 'Driving' and 'Stationary' modes based on the vehicle's state.
The design philosophy is centred on simplifying complex workflows to enhance the user experience. This enables the driver to effortlessly obtain information and execute commands while on the move, ensuring every interaction is intuitive, efficient, and user-focused.
Intuitive Mapping: When parked, the main canvas transitions to a comprehensive 3D vehicle model and status display, allowing control actions to directly correspond to visible objects.
Chunked Browsing: Service cards unfold on the left, enabling faster scanning and management of vehicle status and frequent services (Achieving Glanceability).
Task-Driven Contraction: Upon entering the driving state, the interface actively suppresses secondary information, dedicating the visual focus to navigation and essential feedback.
Low-Interference Strategy: The dark interface and restrained visual hierarchy collectively ensure readability and driving focus under complex lighting conditions.
Core Gesture Interaction: Reducing Visual Reliance
In fast-paced driving environments, the system reduces visual reliance through a tolerance-first action design. By replacing high-precision tapping with broader swiping and direct manipulation (leveraging gross motor skills), interaction safety is significantly enhanced. Additionally, low-friction view switching is achieved via edge-swiping and 3D model manipulation, making cross-scenario focus transitions remarkably smooth.
Service Widget System: Modular Information Gateway
The widget system establishes a highly modular information gateway. High-frequency data and core states are chunked and kept at a surface level, drastically reducing the need for deep navigation. During abnormal vehicle states, the system introduces immediate visual interventions using distinct color contrast (triggering pre-attentive processing) for rapid identification. Furthermore, these adaptable containers preserve flexible space for intuitive sorting and personalized arrangement.
Frequent in-drive vehicle controls require a global entry point that can be called up without breaking the current task. Through a zero-layer pull-down panel, clear grouping logic, and a reconfigurable control layout, Quick Access concentrates core actions into an easier-to-reach zone while balancing reach efficiency with cognitive load.
Icon-First Global Access:
Quick Access organises frequent controls through icon-first recognition, reducing reading demand and interpretation cost while driving. By bringing common actions into a single global layer, adjustments can be accessed and completed within a shorter path.
Cognitive Chunking & Priority Layout:
To reduce cognitive load, controls are organised into compact groups based on function and frequency of use. More critical targets are given clearer size and placement priority, shortening search paths and making repeated interactions easier to internalise over time.
Personalised Control Layout
Driving routines and personal habits vary, which makes fixed control layouts inherently limited. By introducing a reconfigurable grid and component pool, the system allows users to place frequent controls in more accessible positions while maintaining a consistent visual structure. Over time, this helps the control layer align more closely with individual habits and supports faster muscle memory.
As defined earlier in the interaction strategy, ADAS configuration is a typical low-frequency, high-risk task. In this context, the interface needs to support interpretation before action. Through clearer configuration hierarchy, visible state cues, and contextual explanation, the system helps users build more accurate expectations before enabling a function.
Progressive disclosure:
Base assistance, advanced assistance, and safety settings are separated into clearer levels, so users only process the amount of information needed at each step.
Visible system state:
Toggles, modes, and intensity presets stay within the same field of view, making current state and expected behaviour easier to verify together.
Context-linked illustration:
The visual lane model and steering-wheel hints connect software settings to on-road behaviour and physical controls.
Scenario-based explanation:
Functions are explained through driving situations rather than abstract labels alone.
Reduced ambiguity:
Trigger conditions, warning behaviour, and functional boundaries are made easier to grasp through visual explanation.
Deliberate confirmation step:
The confirmation action is supported by clearer expectations, rather than reduced to a purely formal click.
ADAS Status & Alert Visualisations /
Making live system states easier to recognise and differentiate
Moving from configuration to active driving, the system requires a visual language that turns live assistance states into immediate driver awareness. Rather than relying on isolated alerts, the interface uses consistent colour semantics, lane-based visuals, and a clear state hierarchy to organise operational states, cautions, and warnings into a feedback system that is easier to scan at a glance.
Unified status language
Different functions share a consistent visual grammar and colour hierarchy, reducing reliance on label-by-label reading
Severity through visual hierarchy
Normal operation, caution, and warning states are separated through colour and visual intensity, helping users judge the current condition more quickly.
Glanceable feedback system
Lane visuals, surrounding vehicles, and alert cues are integrated into the same field of view, making feedback easier to scan during driving.
Climate and seat adjustments are among the most frequent comfort-related tasks in the cabin. To reduce the effort of reading parameters and locating controls, the interface connects digital controls more directly to cabin space and zones of effect, making adjustments easier to understand, verify, and repeat in everyday use.
Direct spatial mapping:
Airflow direction and area of effect are mapped directly onto the cabin, reducing the need to interpret abstract HVAC icons.
Independent zonal control:
Driver, passenger, and cabin zones remain clearly separated, making adjustments easier to locate, compare, and confirm.
Controls and result in one view:
Temperature, mode, airflow, and environmental change stay within the same field of view, reducing back-and-forth between setting and outcome.
Zone-based mapping:
Heating, ventilation, massage, and steering-wheel functions are organised around their zones of effect, making the target of each adjustment easier to understand.
Separated function and intensity:
Function choice and intensity levels are split into clearer layers, making current settings easier to read.
Integrated comfort layer:
Seat and steering-wheel adjustments stay within one comfort layer, so related controls are not fragmented across different paths.
Charging management requires both fast status reading and ongoing charging decisions. Through clear hierarchy, visual feedback, and scenario-based settings, the interface keeps status, charging progress, and strategy within the same view.
Priority-driven layout:
Range, charging status, and remaining time are placed where they can be read first, with primary actions following naturally.
Tangible charging feedback:
The vehicle silhouette, battery level, and charging metrics communicate change beyond numbers alone.
Scenario-based strategy setting:
Charge cap and slower-charging settings are framed through daily and long-trip contexts, making them easier to judge.
In-car music use includes both content discovery and the ongoing sense of companionship during playback. To balance complex browsing with low visual load while driving, the interface separates library browsing and active playback into two clearer states: one optimised for fast discovery and entry, the other for focused listening and lightweight control.
Recommendation-led visual discovery:
Large recommendation cards occupy the main visual area and rely on cover-led recognition, allowing discovery to depend more on scanning than on active search.
Large targets and shallow playback entry:
Content cards expand the effective touch area and allow playback to begin without entering a deeper detail page, shortening the path from browsing to listening.
Persistent playback anchor:
The mini player remains fixed at the bottom, keeping the current listening state and playback controls available during browsing.
Current content comes into focus:
Album art, track information, and lyrics move to the foreground, shifting the interface from finding content to following what is currently playing.
Controls stay close to the playback state:
Play, skip, favourite, and progress controls are arranged around the main visual, keeping frequent actions accessible without pulling attention away from the content.
Atmosphere changes with the content:
Background tone and record-inspired visuals reinforce the current song, giving the playback screen a more cohesive emotional character beyond pure utility.
Evaluation was conducted through a lightweight mixed-method framework combining heuristic review, cognitive walkthrough, task-based prototype testing, and SUS. Together, these methods were used to revisit the HMI as a whole through interaction efficiency, visual demand, structural clarity, and perceived usability.
Evaluation Framework /
Turning earlier challenges into testable measures
The driving interaction challenges identified earlier were translated into measurable usability criteria at this stage, making it possible to review whether the final interface produced clearer outcomes in efficiency, visual demand, perceived usability, and structural consistency.
Problem
What It Means
Key Metrics
Functionality vs. Simplicity
Are frequent tasks still buried in unnecessarily deep menu structures?
Interaction steps / KLM-estimated task time
Safety vs. Attention
Do key states and controls add unnecessary visual demand while driving?
1.5s recognition / visual target count
Expectation vs. Reality
Is system behaviour clear, predictable, and aligned with user expectations?
SUS score / key understanding gaps
Fragmentation vs. Consistency
Do different modules still behave like separate pages rather than one continuous HMI system?
consistency review / feature discoverability
Key Results/
Reading the final solution through system-level indicators
Findings were consolidated into four system-level indicators covering interaction depth, short-glance recognition, perceived usability, and cross-module consistency.
2
High-frequency core tasks were contained within a 2-step path, with a KLM-estimated task time of 1.8s.
Frequent actions were brought back to a shallower layer, reducing search, backtracking, and error recovery.
Related to / Functionality vs. Simplicity
95%
Critical states reached 95% recognition accuracy under a 1.5-second exposure check.
Safety-related information no longer depends on full reading, and supports shorter visual dwell.
Related to / Safety vs. Attention
82.5
The system reached a SUS score of 82.5 / 100, while 4 key understanding gaps were resolved.
Spatial mapping, contextual explanation, and configurable access brought system behaviour closer to user expectations.
Related to / Expectation vs. Reality
12 / 12
All 12 core criteria in the cross-module consistency review were met.
Modules can now be read as one continuous, predictable HMI language rather than isolated feature pages.
Related to / Fragmentation vs. Consistency
Design Implications /
Judgements further clarified through evaluation
The evaluation does not replace in-vehicle testing, but it sharpened several design judgements: frequent tasks depend on shallower entry points, low-frequency safety-critical functions depend more on explanation quality, and cross-module consistency directly shapes long-term use.
For short-duration tasks such as temperature adjustment, playback control, and quick toggles, reducing path depth still matters more than adding functions.
Improvements in ADAS-related tasks came more from state hierarchy, explanation, and confirmation logic than from simply removing a step.
When navigation, media, vehicle control, and settings share one structural and state grammar, users are more likely to understand them as one HMI rather than separate modules.
Next step Further validation should focus on short-glance behaviour under real driving workload, error recovery cost under time pressure, and cross-screen coordination across the cluster, HUD, and steering-wheel controls. This would make it possible to evaluate attention distribution across the broader driving interface ecosystem, not only the centre display.
