结论：手感不是玄学，它由几个可测的参数决定
使用者对轮椅操控的评价常常是主观描述：「反应太灵敏」「起步太冲」「停不准」。这些感受背后对应的是几个具体参数，都可以测量和调整。
把主观描述对应到参数上，改进才有方向，否则只能凭感觉反复调，调了也说不清改善在哪。
该测哪些特性
| 特性 | 影响什么 | 测量方法 |
|---|---|---|
| 死区大小 | 手抖时会不会误动作 | 逐渐增大摇杆位移，记录开始输出的位置 |
| 位移与输出的关系 | 操控的线性感 | 记录不同位移下的输出，画特性曲线 |
| 加速响应 | 起步是否突兀 | 摇杆阶跃输入，记录速度上升过程 |
| 减速响应 | 停车是否准确 | 摇杆回中，记录速度下降过程 |
| 回中精度 | 松手后会不会漂移 | 反复操作后测量回中位置 |
| 方向耦合 | 直行时会不会跑偏 | 纯前进输入下测量转向输出 |
| 操作力 | 使用者能否操作 | 测量各方向的操作力 |
死区的取舍
死区是摇杆在中位附近的不响应区间。它的作用是防止轻微抖动或碰触造成误动作。
死区大，安全性好但操控迟钝；死区小，响应灵敏但容易误触发。 合适的值取决于目标使用者：手部稳定性较差的使用者需要较大死区，而操作精细的使用者会觉得大死区难用。
有些产品把死区做成可调参数，由配置人员按使用者情况设定。这是合理的设计，但要注意：可调就意味着可能被调到不合适的值，所以调节范围本身应当有限制，并且配置过程应当有记录。
特性曲线的形状
位移与输出之间的关系可以是线性的，也可以是非线性的。
线性关系直观，但在低速精细操作时分辨率不足——摇杆推一点点，速度变化已经不小。非线性关系（低位移段斜率小、高位移段斜率大）能兼顾低速精细和高速效率，但使用者需要适应。
这个选择应当基于使用场景：室内为主、需要频繁精细操作的，倾向非线性；户外为主、以移动效率为先的，线性更直观。
加减速响应的验证
加速响应测的是摇杆推到位之后，速度上升到目标值需要多久、过程是否平滑。这个参数与稳定性直接相关——加速过快会增加后倾风险。
减速响应测的是摇杆回中之后的减速过程。这里有个实际问题：减速太快，使用者会前倾不适；减速太慢，停不准且有碰撞风险。
两者的设定应当与稳定性试验的结果一起考虑，不能只看手感。测出来手感好的设定，如果在稳定性试验中表现不佳，还是要调整。
方向耦合的检查
方向耦合是指纯前进的输入却产生了转向输出，表现为车辆跑偏。原因可能是摇杆机械结构的不对称、电位器的一致性、或者软件处理的偏差。
这一项在实验室容易测，但在产品定义里常常没有指标。建议自己设一个内部指标，因为跑偏是使用者感知很明显的问题，而且在长距离行驶中会累积成明显偏差。
特殊控制装置的处理
部分使用者无法操作标准摇杆，需要其他控制方式：头控、吹吸控制、下颌控制、扫描式开关等。
这些装置的评价原则相同——死区、响应、误动作防护——但具体参数差别很大。例如吹吸控制的「死区」是压力阈值，头控的误动作风险来自乘坐颠簸时头部的非意图移动。
这类装置的验证要特别关注非意图输入的防护，因为它们的输入信号来源更容易被环境干扰。
参数与使用者能力的匹配
控制器参数不是定死的，而是应当按使用者配置。这带来一个体系上的要求：配置过程要有记录，配置人员要有相应能力。
对制造商而言，需要做的是：规定各参数的允许范围；提供配置工具与指导；在说明书中说明各参数的作用与风险。范围的边界应当由试验确定——比如加速度调到范围上限时，稳定性试验仍然合格。
与使用者培训的衔接
摇杆参数配置好之后，使用者需要适应。适应期内误操作的风险较高，尤其是参数改动较大时。
建议在产品交付流程里包含试用与调整环节：先按保守参数交付，使用者熟悉之后再按需要调整。一步到位调到灵敏设定，对新使用者是有风险的。
说明书中也应当提示：参数调整后应当在安全环境下试用，确认适应后再在复杂环境中使用。
控制装置的耐久
摇杆是高频操作部件，耐久同样需要考察。关注点包括：反复操作后的回中精度变化、电位器或传感器的磨损、防水密封的老化、以及操作力的变化。
回中精度的劣化是最需要关注的一项，因为它直接导致松手后车辆漂移，而使用者未必能及时察觉。
环境因素对响应的影响
摇杆的响应特性会受环境影响：低温下润滑脂变稠，操作力上升；高温或潮湿可能影响传感元件的输出；长期紫外照射会让弹性件老化，回中力下降。
如果产品面向温差较大的地区，建议在温度边界条件下补测响应特性，确认在极端条件下仍在可接受范围内。这项在常规清单里通常没有，但对户外产品有实际意义。
多人共用时的参数管理
在养老机构、康复中心这类场景，同一台轮椅可能由不同使用者轮流使用，而每个人适合的参数不同。
这带来管理需求：参数配置要能快速切换并可追溯，避免上一个使用者的设定被下一个使用者沿用。有些产品支持多组参数预设，这在机构场景是有价值的功能，但要确保切换过程不会误选。
我们的做法
做摇杆特性测量时，我们会输出特性曲线而不只给几个数值。曲线能看出数值看不出的问题，比如某个位移区间的输出突变、回中附近的不连续。
对于参数可调的产品，建议在参数范围的边界值上补做稳定性和制动验证，确认整个可调范围内产品都是安全的。只在出厂默认值上做验证，覆盖不到使用者实际使用的设定。
有需要可以把控制器的参数范围和目标使用者特征发过来一起定测量方案，或者直接联系：132 4819 8029。检测能力见服务介绍，产品分类见电动轮椅检测与代步车检测，联系方式见联系我们。
English version
Conclusion. Users describe wheelchair control subjectively: too sensitive, too abrupt off the mark, hard to stop accurately. Behind those impressions sit a handful of specific parameters, all measurable and adjustable. Mapping the subjective description onto parameters is what gives improvement a direction; without it, tuning proceeds by feel and the result cannot be explained.
What to measure. Deadband size determines whether tremor causes unintended movement, measured by increasing joystick displacement until output begins. The relationship between displacement and output determines the linearity of control, measured by recording output across the range to produce a characteristic curve. Acceleration response determines whether starting is abrupt, measured with a step input while recording the speed rise. Deceleration response determines stopping accuracy, measured by returning the joystick to centre and recording the speed fall. Centring accuracy determines whether the chair drifts when the joystick is released, measured after repeated operation. Directional coupling determines whether the chair pulls to one side, measured as steering output under a pure forward input. Operating force determines whether the user can operate it at all.
The deadband trade-off. The deadband is the unresponsive region near centre, guarding against tremor or accidental contact. A large deadband improves safety but dulls response; a small one gives sensitivity but triggers easily. The right value depends on the intended user: someone with limited hand stability needs a larger deadband, while a user capable of fine control finds a large one awkward. Some products make it configurable, which is reasonable, but configurable means it can be set inappropriately, so the adjustment range itself should be bounded and configuration should be recorded.
Shape of the characteristic curve. The displacement-to-output relationship may be linear or not. Linear is intuitive but offers poor resolution for slow precise manoeuvring, since a small joystick movement already produces a noticeable speed change. A non-linear curve, shallow near centre and steeper further out, serves both precision and efficiency but requires the user to adapt. The choice should follow the use scenario: indoor use with frequent fine manoeuvring favours non-linear, outdoor use prioritising travel favours linear.
Acceleration and deceleration. Acceleration response measures how long speed takes to reach target after a full input and whether the rise is smooth, and it connects directly to stability, since rapid acceleration increases rearward tipping risk. Deceleration response measures the slowing after the joystick returns to centre, where too fast is uncomfortable and throws the user forward while too slow means imprecise stopping and collision risk. Both settings should be considered alongside stability results rather than by feel alone: a setting that feels good but performs poorly in stability testing still has to change.
Directional coupling. This is steering output arising from a purely forward input, seen as the chair pulling to one side. Causes include mechanical asymmetry in the joystick, potentiometer matching, or a software offset. It is easy to measure in the laboratory but is rarely given a specification. Setting an internal indicator is worthwhile, because users notice pulling readily and it accumulates into a marked deviation over distance.
Alternative control devices. Some users cannot operate a standard joystick and use head control, sip-and-puff, chin control or scanning switches. The evaluation principles are the same, covering deadband, response and protection against unintended input, but parameters differ greatly: the deadband of a sip-and-puff device is a pressure threshold, and the unintended-input risk for head control comes from involuntary head movement over rough ground. Protection against unintended input deserves particular attention with these devices, because their input signals are more readily disturbed by the environment.
Matching parameters to user capability. Controller parameters are configured per user, which imposes a system requirement: configuration must be recorded and configuring personnel must be competent. The manufacturer must define the permitted range of each parameter, supply configuration tools and guidance, and explain in the instructions what each parameter does and what risk it carries. The boundaries of the range should be established by testing: with acceleration at the top of its range, stability testing must still pass.
How we handle it. We produce characteristic curves rather than a few numbers, because a curve reveals things numbers do not, such as an output discontinuity in part of the travel or irregularity near centre. For products with adjustable parameters we suggest supplementary stability and braking verification at the range limits, confirming the product is safe across the whole adjustable range rather than only at the factory default.
Send us the parameter ranges and a description of the intended users and we will design the measurement. Phone or WeChat: +86 132 4819 8029.