An a priori power analysis was performed with G*Power 3.1 (two-tailed, independent-samples t-test; allocation ratio 1:1). To buffer a projected 10% attrition, we aimed for 24 enrolments (12 per group). Logistical constraints—including academic-calendar limits and restricted headset inventory—ultimately confined recruitment to 20 participants (10 per arm). We mitigated this limitation by: (i) reporting all outcomes with 95% confidence intervals and standardized effect sizes; (ii) supplementing primary tests with Bayesian estimation to quantify evidence strength; and (iii) highlighting power constraints in the discussion, positioning this single-centre trial as a hypothesis-generating precursor to an upcoming multi-site study.

In the game, children can choose their favorite animals to raise, give them names, and assign tasks such as taking a bath, going for a walk, having a meal, or going to sleep. At the bottom of the interface, they can enter a numerical value to specify the length of the activity. During this process, the animal will communicate with children in real time, provide feedback, and create a cordial atmosphere, thus aiding in the therapy for autism. The animals will make friends during their walks in the wild and they will bring their friends to their homes. This design encourages autistic children to open up, overcome their fears, make friends, and improve their social skills. Energy is consumed when children assign tasks. When the energy level is low, children need to pedal a stationary bicycle to replenish energy in the real world, as shown in Figure 1. This is a perfect combination of cycling sports and games and it is scientifically proven that exercise makes children happy and is more beneficial to the rehabilitation of autistic children (19).

The game can be operated with or without VR headsets and is compatible with computers, smartphones, and other mobile devices. The procedure is as follows:

Peripheral Devices: A cadence-sensor bicycle with ANT+/Bluetooth transceiver (US $120) supplies real-time power output; any alternative stationary bike can be integrated via a serial-COM bridge. For motion tracking without head-mounted display (HMD) the system recognises keyboard, gamepad, or touch input, ensuring accessibility when VR optics are unaffordable.

Software Stack: Unity 2021.3 LTS, OpenXR runtime, and SQLite local database; Python 3.10 scripts perform data logging and anonymised batch export. The build operates fully offline and synchronises via Wi-Fi 802.11n when connectivity permits, satisfying General Data Protection Regulation (GDPR) and local privacy statutes.

Adaptations for Low-Resource Settings: Opt-in “2D-lite” mode disables high-poly assets and substitutes prerecorded 720p video loops, reducing GPU load ≈ 60%.

Text strings and voice-overs are externalized in JSON, enabling rapid localization. All source code and 3-D assets (except licensed music) are released under CC-BY-NC 4.0 on GitHub; hardware schematics for the bike-interface PCB are included to encourage local fabrication.

These specifications align with recent VR-CORE recommendations for clinical trials in immersive environments, facilitating replication and scalability across diverse socioeconomic contexts.

In the realm of game design, the conceptualization of systemic design strategies is crucial to the game’s efficacy. Based on Raaja Shri (20) research theory and practical experience, combined with our actual research, we have concluded that there are three main aspects of game design. First, provide information presentation that fits the autistic children’s personality thinking, which can emphasize the system’s ability to create visual and auditory interventions for the patient. Second, to create a safe intervention environment for autistic children, compared with normal children, autistic children have speech and behavior disorders. Virtual reality games should consider the behavioral characteristics of patients and create a safe and efficient intervention environment. Third, provide personalized treatment for autistic children, simple game content models were summarized that could provide a wide range of treatments for children with autism. The intervention is deliberately designed so that children with autism need neither prior technical literacy to interact with the game nor any previous gaming experience during the experimental sessions.

From the above three aspects lead to the following principles for designing serious games that intervene in the rehabilitation therapy of children with autism, as shown in Figure 2.

The game is structured into five thematic “quest clusters” (Self-Care, Animal Bonding, Outdoor Exploration, Cooperative Tasks, Emotion Labeling).

Tiered Complexity: Each cluster contains three tiers. Tier-1 tasks are single-step (e.g., “feed the panda 200 kcal”), Tier-2 add social cues (virtual friend requests), Tier-3 demand joint attention (simultaneous actions with an NPC).

Dynamic Adjustment: A Bayesian learner monitors real-time performance (error rate, task latency) and modulates (i) visual clutter, (ii) NPC conversational turns, and (iii) required pedaling watts.

Flow Maintenance: Difficulty is increased when success > 80% over three consecutive trials and decreased when < 40%.

Intrinsic Motivation: Token economy (virtual badges convertible to print-out stickers) and avatar personalisation sustain engagement beyond novelty fade.

The development of the interactive rehabilitation game was an intricate process that involved a multidisciplinary approach. The game’s design was rooted in the psychological and physiological needs of children with Autism Spectrum Disorder (ASD), aiming to provide a therapeutic experience through immersive technology. The process commenced with a thorough analysis of child psychology, followed by the application of game based intervention strategies. The game was designed to leverage the principles of flow theory to enhance user engagement and therapeutic outcomes.

The development team was composed of experts from various fields, including multimedia and film media art, imaging science, psychology, game design, and education. Collaborative efforts were essential, involving specialists in ASD therapy, educational psychologists, and technical professionals in virtual reality (VR) and motion-sensing technology.

The game was constructed using the Unity3D engine, which facilitated the creation of a three-dimensional (3D) interactive environment. The development process involved the following stages:

The game was designed to be compatible with a range of VR headsets, specifically the Hypereal VR HMD, to provide an immersive visual experience. Additional equipment included:

The exercise bike was interfaced with the game through a serial port communication protocol. The bike’s electronics captured physical data such as pedaling speed, heart rate, and cycling distance. This data was packaged into a data stream and transmitted to the game software via a USB connection. The game software featured a script that continuously polled the serial port for incoming data, parsed the stream, and converted it into in game energy points. This energy was then used to power various activities within the game, creating a closed loop of physical exertion and virtual reward. The system was designed to ensure that the child’s physical activity levels influenced their progress and interactions within the game environment.

When assumptions were breached, Mann–Whitney U (between-group) or Wilcoxon signed-rank (within-group) tests supplanted parametric analyses. Exact two-tailed p-values and rank-biserial effect sizes are reported alongside 95% CIs. All analyses adhered to CONSORT 2010 and utilized SPSS v20.0, for the processing and analysis of the study’s data. [N(%)] is used to represent count data and X 2 test and Fisher exact test are used for comparison and analysis among groups. ( { x ¯ ± δ ) is used to indicate measurement data and comparison and analysis among groups are performed by using t-test. (When the difference has a statistical significance, p < 0.05.)

Primary endpoint: Change in Social Responsiveness Scale-2 (SRS-2) total score from baseline (T0) to end-of-treatment (T1, Week 16). The SRS-2 is a validated, caregiver-rated instrument for autism-related social impairment.

Secondary endpoints:

Assessment schedule: T0 = baseline; T1 = Week 16; T2 = follow-up Week 24. All psychometric testing is administered by blinded assessors who are independent of the intervention team.

Statistical analysis plan: All analyses follow an intention-to-treat approach. A linear mixed-effects model (fixed factors = group, time; random = subject ID) estimates group × time interactions. Post-hoc contrasts employ Bonferroni correction. Effect sizes are reported as Hedges’ g with 95% CIs. Missing data are multiply imputed (m = 20) using predictive mean matching. Bayesian parameter estimation (weakly informative priors) supplements frequentist inference to quantify evidence strength. Adverse events—including Simulator Sickness Questionnaire scores ≥ 15—are summarized descriptively.

Upon examination of Table 2, it is evident that there is a notable difference in the SDS and SAS scale assessment scores between children with autism who engaged in the interactive VR-Motion serious game intervention ( Y age =119.4 months SD = 29.41 months) and those who did not partake in the intervention. Compared with those who do not participate in the interactive VR-Motion serious game intervention. Among them, the scores of autistic children who have carried out the game intervention are significantly lower than those who have not carried out the game intervention.

Through three groups of experiments, the valid measured data of 19 autistic children are analyzed. It can be seen that after the use of immersive VR game to participate in the therapy, the effect has a significant difference. Among the three groups of data, experimental group 2 has the best improvement effect (using 16 weeks of VR) while the improvement effect of control group 1 (without the serious game) is significantly lower than that of the remaining groups. The data analysis of SDS and SAS scales at the end of the experiment is shown in Table 2.

The experimental data analysis shows that the design of this interactive VR-Motion serious game is relatively reasonable and it has a good effect on the development of autistic children’s social communication and self-care ability. In the post-test of wish tracking, df = 9; t = −1.155; p = 0.281, which can be seen that there is a good improvement in social intentions of autistic children after carrying out an immersive VR game experience. The experts indicate in their opinions that the design of this immersive VR game has superior guidance for simulating trial social contact and experiencing warm social emotions of children with autism spectrum disorder (ASD) and has excellent benefits in cultivating their independent willingness to participate in social contact and introducing them into life situations correctly in the game process.

The intervention treatment of experimental interactive VR-Motion serious game for autistic children through pre-experiment and formal experiment has a better improvement effect on alleviating the depression and anxiety of ASD children. By comparing the experimental SDS and SAS levels, the recovery degree of levels is better and higher after the game is intervened, which is enough to see that the positive effect in social skills and social life is substantial (see Table 2). Thus, it can prove my presupposition that this interactive VR-Motion serious game can play a positive role in the application of treating autistic children.

Since the causes of ASD and the influence factors are the combined action of multiple factors, this creates a lot of inconveniences and difficulties in the treatment of autistic children. After taking two experiments, preliminary phase conclusions have been drawn through the analysis of the data collected in the intervention treatment acting on autistic children and the results obtained from callback effect analysis of a part of selected autistic children in the experimental group are also very optimistic, as shown in Table 3.

It can be clearly seen from the Figure 3 that under the interactive VR-Motion serious game, the unstable emotional level of autistic children is gradually decreasing overall. The average unstable emotional level of the autistic children in the experimental group improved better after 14 weeks than the control group(traditional treatment) who do not receive the serious game therapy. Thus, it can be seen that this game is designed to have a significant impact on practical applications for autistic children and can be added to or used in cooperation with regular psychotherapy for autistic children, which has an excellent effect on improving their social contact willingness and life skills and stabilizing their emotions.

Younger children with autism often react to environmental stimuli, leading to emotional stress when introduced to new environments such as a gaming studio. During our study, subjects under the age of 10 exhibited varying levels of emotional stress when first exposed to the serious game environment. To mitigate this, we included a familiarization period where the children could explore the game environment at their own pace before the formal intervention began. Additionally, familiar caregivers were present during initial sessions to help reduce anxiety and emotional stress. The enforcement of the experiment for subjects under the age of 10 required special attention. Extra care was taken to ensure that these younger children were comfortable and engaged. For instance, breaks were provided more frequently, and the duration of gaming sessions was adjusted based on the child’s comfort level and attention span. This approach helped maintain a supportive and non-threatening environment, which is essential for the well-being and effective participation of younger subjects.

While the interactive VR-Motion serious game showed promising results, there are several drawbacks that need to be addressed for future improvements:

This study has several constraints. First, the underpowered sample (N < 20) increases susceptibility to Type II error and restricts subgroup analyses. Second, the trial was conducted in a single metropolitan centre, limiting ecological and cultural generalisability. Third, intervention fidelity depended on bespoke VR hardware (~US$ 600 per unit), which may impede scalability in low-resource settings. Fourth, self-report instruments (SDS/SAS) are vulnerable to social-desirability bias, although we mitigated this by blinded administration. Finally, the absence of a long-term follow-up (> 6 months) precludes definitive conclusions on sustained efficacy. Future multisite trials with cost-effectiveness analyses are warranted.