diff --git a/libuavcan/include/uavcan/dynamic_memory.hpp b/libuavcan/include/uavcan/dynamic_memory.hpp
index 927b8f392a..ff0ef2af14 100644
--- a/libuavcan/include/uavcan/dynamic_memory.hpp
+++ b/libuavcan/include/uavcan/dynamic_memory.hpp
@@ -7,7 +7,6 @@
 #include <cassert>
 #include <cstdlib>
 #include <cstring>
-#include <limits>
 #include <uavcan/stdint.hpp>
 #include <uavcan/util/templates.hpp>
 #include <uavcan/impl_constants.hpp>
@@ -41,8 +40,8 @@ class UAVCAN_EXPORT PoolManager : public IPoolAllocator, Noncopyable
     {
         const IPoolAllocator* const a = *static_cast<const IPoolAllocator* const*>(raw_a);
         const IPoolAllocator* const b = *static_cast<const IPoolAllocator* const*>(raw_b);
-        const std::size_t a_size = a ? a->getBlockSize() : std::numeric_limits<std::size_t>::max();
-        const std::size_t b_size = b ? b->getBlockSize() : std::numeric_limits<std::size_t>::max();
+        const std::size_t a_size = a ? a->getBlockSize() : NumericTraits<std::size_t>::max();
+        const std::size_t b_size = b ? b->getBlockSize() : NumericTraits<std::size_t>::max();
         if (a_size != b_size)
         {
             return (a_size > b_size) ? 1 : -1;
diff --git a/libuavcan/include/uavcan/marshal/float_spec.hpp b/libuavcan/include/uavcan/marshal/float_spec.hpp
index 7c110f896f..ece754b39a 100644
--- a/libuavcan/include/uavcan/marshal/float_spec.hpp
+++ b/libuavcan/include/uavcan/marshal/float_spec.hpp
@@ -4,7 +4,6 @@
 
 #pragma once
 
-#include <limits>
 #include <cmath>
 #include <uavcan/stdint.hpp>
 #include <uavcan/data_type.hpp>
@@ -18,6 +17,8 @@
 #endif
 #if UAVCAN_CPP_VERSION < UAVCAN_CPP11
 # include <math.h>     // Needed for isfinite()
+#else
+# include <limits>     // Assuming that in C++11 mode all standard headers are available
 #endif
 
 namespace uavcan
@@ -43,8 +44,10 @@ class UAVCAN_EXPORT IEEE754Converter
     IEEE754Converter();
 
 public:
+#if UAVCAN_CPP_VERSION >= UAVCAN_CPP11
     /// UAVCAN requires rounding to nearest for all float conversions
     static std::float_round_style roundstyle() { return std::round_to_nearest; }
+#endif
 
     template <unsigned BitLen>
     static typename IntegerSpec<BitLen, SignednessUnsigned, CastModeTruncate>::StorageType
@@ -118,11 +121,17 @@ public:
 
     typedef typename NativeFloatSelector<BitLen>::Type StorageType;
 
+#if UAVCAN_CPP_VERSION < UAVCAN_CPP11
+    enum { IsExactRepresentation = (sizeof(StorageType) * 8 == BitLen) };
+#else
     enum { IsExactRepresentation = (sizeof(StorageType) * 8 == BitLen) && std::numeric_limits<StorageType>::is_iec559 };
+#endif
 
     using IEEE754Limits<BitLen>::max;
     using IEEE754Limits<BitLen>::epsilon;
+#if UAVCAN_CPP_VERSION >= UAVCAN_CPP11
     static std::float_round_style roundstyle() { return IEEE754Converter::roundstyle(); }
+#endif
 
     static int encode(StorageType value, ScalarCodec& codec, TailArrayOptimizationMode)
     {
@@ -180,11 +189,11 @@ private:
         {
             if (value > max())
             {
-                value = std::numeric_limits<StorageType>::infinity();
+                value = NumericTraits<StorageType>::infinity();
             }
             else if (value < -max())
             {
-                value = -std::numeric_limits<StorageType>::infinity();
+                value = -NumericTraits<StorageType>::infinity();
             }
             else
             {
diff --git a/libuavcan/include/uavcan/marshal/integer_spec.hpp b/libuavcan/include/uavcan/marshal/integer_spec.hpp
index a0a6e298e1..ab6ea98028 100644
--- a/libuavcan/include/uavcan/marshal/integer_spec.hpp
+++ b/libuavcan/include/uavcan/marshal/integer_spec.hpp
@@ -4,7 +4,6 @@
 
 #pragma once
 
-#include <limits>
 #include <uavcan/stdint.hpp>
 #include <uavcan/data_type.hpp>
 #include <uavcan/util/templates.hpp>
@@ -93,9 +92,9 @@ private:
     {
         StaticAssert<(BitLen <= (sizeof(StorageType) * 8))>::check();
         // coverity[result_independent_of_operands : FALSE]
-        assert(max() <= std::numeric_limits<StorageType>::max());
+        assert(max() <= NumericTraits<StorageType>::max());
         // coverity[result_independent_of_operands : FALSE]
-        assert(min() >= std::numeric_limits<StorageType>::min());
+        assert(min() >= NumericTraits<StorageType>::min());
     }
 
 public:
diff --git a/libuavcan/include/uavcan/marshal/scalar_codec.hpp b/libuavcan/include/uavcan/marshal/scalar_codec.hpp
index 2b7bceb0eb..be0aa373ba 100644
--- a/libuavcan/include/uavcan/marshal/scalar_codec.hpp
+++ b/libuavcan/include/uavcan/marshal/scalar_codec.hpp
@@ -5,7 +5,6 @@
 #pragma once
 
 #include <cassert>
-#include <limits>
 #include <uavcan/stdint.hpp>
 #include <uavcan/impl_constants.hpp>
 #include <uavcan/util/templates.hpp>
@@ -47,10 +46,10 @@ class UAVCAN_EXPORT ScalarCodec
     convertByteOrder(uint8_t (&)[Size]) { }
 
     template <unsigned BitLen, typename T>
-    static typename EnableIf<std::numeric_limits<T>::is_signed && ((sizeof(T) * 8) > BitLen)>::Type
+    static typename EnableIf<static_cast<bool>(NumericTraits<T>::IsSigned) && ((sizeof(T) * 8) > BitLen)>::Type
     fixTwosComplement(T& value)
     {
-        StaticAssert<std::numeric_limits<T>::is_integer>::check(); // Not applicable to floating point types
+        StaticAssert<NumericTraits<T>::IsInteger>::check(); // Not applicable to floating point types
         if (value & (T(1) << (BitLen - 1)))                        // The most significant bit is set --> negative
         {
             value |= 0xFFFFFFFFFFFFFFFF & ~((T(1) << BitLen) - 1);
@@ -58,7 +57,7 @@ class UAVCAN_EXPORT ScalarCodec
     }
 
     template <unsigned BitLen, typename T>
-    static typename EnableIf<!std::numeric_limits<T>::is_signed || ((sizeof(T) * 8) == BitLen)>::Type
+    static typename EnableIf<!static_cast<bool>(NumericTraits<T>::IsSigned) || ((sizeof(T) * 8) == BitLen)>::Type
     fixTwosComplement(T&) { }
 
     template <unsigned BitLen, typename T>
@@ -77,7 +76,7 @@ class UAVCAN_EXPORT ScalarCodec
     {
         StaticAssert<((sizeof(T) * 8) >= BitLen)>::check();
         StaticAssert<(BitLen <= BitStream::MaxBitsPerRW)>::check();
-        StaticAssert<std::numeric_limits<T>::is_signed ? (BitLen > 1) : 1>::check();
+        StaticAssert<static_cast<bool>(NumericTraits<T>::IsSigned) ? (BitLen > 1) : true>::check();
     }
 
     int encodeBytesImpl(uint8_t* bytes, unsigned bitlen);
diff --git a/libuavcan/include/uavcan/time.hpp b/libuavcan/include/uavcan/time.hpp
index 68c7b51286..3f25f3448e 100644
--- a/libuavcan/include/uavcan/time.hpp
+++ b/libuavcan/include/uavcan/time.hpp
@@ -4,7 +4,6 @@
 
 #pragma once
 
-#include <limits>
 #include <uavcan/stdint.hpp>
 #include <uavcan/impl_constants.hpp>
 #include <uavcan/util/templates.hpp>
@@ -29,7 +28,7 @@ protected:
     }
 
 public:
-    static D getInfinite() { return fromUSec(std::numeric_limits<int64_t>::max()); }
+    static D getInfinite() { return fromUSec(NumericTraits<int64_t>::max()); }
 
     static D fromUSec(int64_t us)
     {
@@ -106,7 +105,7 @@ protected:
     }
 
 public:
-    static T getMax() { return fromUSec(std::numeric_limits<uint64_t>::max()); }
+    static T getMax() { return fromUSec(NumericTraits<uint64_t>::max()); }
 
     static T fromUSec(uint64_t us)
     {
@@ -142,7 +141,7 @@ public:
         {
             if (uint64_t(usec_ + r.toUSec()) < usec_)
             {
-                return fromUSec(std::numeric_limits<uint64_t>::max());
+                return fromUSec(NumericTraits<uint64_t>::max());
             }
         }
         return fromUSec(usec_ + r.toUSec());
diff --git a/libuavcan/include/uavcan/transport/transfer_buffer.hpp b/libuavcan/include/uavcan/transport/transfer_buffer.hpp
index 7969338b82..db61104422 100644
--- a/libuavcan/include/uavcan/transport/transfer_buffer.hpp
+++ b/libuavcan/include/uavcan/transport/transfer_buffer.hpp
@@ -4,7 +4,6 @@
 
 #pragma once
 
-#include <limits>
 #include <uavcan/stdint.hpp>
 #include <uavcan/error.hpp>
 #include <uavcan/transport/frame.hpp>
diff --git a/libuavcan/include/uavcan/util/templates.hpp b/libuavcan/include/uavcan/util/templates.hpp
index 438f426be3..eaf9aa354b 100644
--- a/libuavcan/include/uavcan/util/templates.hpp
+++ b/libuavcan/include/uavcan/util/templates.hpp
@@ -4,7 +4,19 @@
 
 #pragma once
 
+#include <climits>
+#include <cmath>
 #include <uavcan/impl_constants.hpp>
+#include <uavcan/stdint.hpp>
+
+#ifndef UAVCAN_CPP_VERSION
+# error UAVCAN_CPP_VERSION
+#endif
+#if UAVCAN_CPP_VERSION < UAVCAN_CPP11
+# include <float.h>      // cfloat may not be available
+#else
+# include <cfloat>       // C++11 mode assumes that all standard headers are available
+#endif
 
 namespace uavcan
 {
@@ -241,4 +253,115 @@ bool equal(InputIt1 first1, InputIt1 last1, InputIt2 first2)
     return true;
 }
 
+/**
+ * Numeric traits, like std::numeric_limits<>
+ */
+template <typename T>
+struct UAVCAN_EXPORT NumericTraits;
+
+/// 8 bit
+template <>
+struct UAVCAN_EXPORT NumericTraits<int8_t>
+{
+    enum { IsSigned = 1 };
+    enum { IsInteger = 1 };
+    static int8_t max() { return 127; }
+    static int8_t min() { return -max() - 1; }
+};
+template <>
+struct UAVCAN_EXPORT NumericTraits<uint8_t>
+{
+    enum { IsSigned = 0 };
+    enum { IsInteger = 1 };
+    static uint8_t max() { return 0xFFU; }
+    static uint8_t min() { return 0; }
+};
+
+/// 16 bit
+template <>
+struct UAVCAN_EXPORT NumericTraits<int16_t>
+{
+    enum { IsSigned = 1 };
+    enum { IsInteger = 1 };
+    static int16_t max() { return 32767; }
+    static int16_t min() { return -max() - 1; }
+};
+template <>
+struct UAVCAN_EXPORT NumericTraits<uint16_t>
+{
+    enum { IsSigned = 0 };
+    enum { IsInteger = 1 };
+    static uint16_t max() { return 0xFFFFU; }
+    static uint16_t min() { return 0; }
+};
+
+/// 32 bit
+template <>
+struct UAVCAN_EXPORT NumericTraits<int32_t>
+{
+    enum { IsSigned = 1 };
+    enum { IsInteger = 1 };
+    static int32_t max() { return 2147483647L; }
+    static int32_t min() { return -max() - 1; }
+};
+template <>
+struct UAVCAN_EXPORT NumericTraits<uint32_t>
+{
+    enum { IsSigned = 0 };
+    enum { IsInteger = 1 };
+    static uint32_t max() { return 0xFFFFFFFFUL; }
+    static uint32_t min() { return 0UL; }
+};
+
+/// 64 bit
+template <>
+struct UAVCAN_EXPORT NumericTraits<int64_t>
+{
+    enum { IsSigned = 1 };
+    enum { IsInteger = 1 };
+    static int64_t max() { return 9223372036854775807LL; }
+    static int64_t min() { return -max() - 1; }
+};
+template <>
+struct UAVCAN_EXPORT NumericTraits<uint64_t>
+{
+    enum { IsSigned = 0 };
+    enum { IsInteger = 1 };
+    static uint64_t max() { return 0xFFFFFFFFFFFFFFFFULL; }
+    static uint64_t min() { return 0; }
+};
+
+/// float
+template <>
+struct UAVCAN_EXPORT NumericTraits<float>
+{
+    enum { IsSigned = 1 };
+    enum { IsInteger = 0 };
+    static float max() { return FLT_MAX; }
+    static float min() { return FLT_MIN; }
+    static float infinity() { return INFINITY; }
+};
+
+/// double
+template <>
+struct UAVCAN_EXPORT NumericTraits<double>
+{
+    enum { IsSigned = 1 };
+    enum { IsInteger = 0 };
+    static double max() { return DBL_MAX; }
+    static double min() { return DBL_MIN; }
+    static double infinity() { return static_cast<double>(INFINITY) * static_cast<double>(INFINITY); }
+};
+
+/// long double
+template <>
+struct UAVCAN_EXPORT NumericTraits<long double>
+{
+    enum { IsSigned = 1 };
+    enum { IsInteger = 0 };
+    static long double max() { return LDBL_MAX; }
+    static long double min() { return LDBL_MIN; }
+    static long double infinity() { return static_cast<long double>(INFINITY) * static_cast<long double>(INFINITY); }
+};
+
 }
diff --git a/libuavcan/src/marshal/uc_float_spec.cpp b/libuavcan/src/marshal/uc_float_spec.cpp
index bc2a6921e5..644dd16f4f 100644
--- a/libuavcan/src/marshal/uc_float_spec.cpp
+++ b/libuavcan/src/marshal/uc_float_spec.cpp
@@ -10,6 +10,10 @@
 # error UAVCAN_CPP_VERSION
 #endif
 
+#if UAVCAN_CPP_VERSION >= UAVCAN_CPP11
+# include <limits>
+#endif
+
 #undef signbit
 #undef isnan
 #undef isinf
@@ -47,7 +51,7 @@ static inline bool isinf(T arg)
 #if UAVCAN_CPP_VERSION >= UAVCAN_CPP11
     return std::isinf(arg);
 #else
-    return arg == std::numeric_limits<T>::infinity() || arg == -std::numeric_limits<T>::infinity();
+    return arg == NumericTraits<T>::infinity() || arg == -NumericTraits<T>::infinity();
 #endif
 }
 
@@ -94,12 +98,20 @@ float IEEE754Converter::halfToNativeNonIeee(uint16_t value)
     unsigned abs = value & 0x7FFFU;
     if (abs > 0x7C00U)
     {
+#if UAVCAN_CPP_VERSION >= UAVCAN_CPP11
         out = std::numeric_limits<float>::has_quiet_NaN ? std::numeric_limits<float>::quiet_NaN() : 0.0F;
+#else
+        out = nanf("");
+#endif
     }
     else if (abs == 0x7C00U)
     {
+#if UAVCAN_CPP_VERSION >= UAVCAN_CPP11
         out = std::numeric_limits<float>::has_infinity ?
               std::numeric_limits<float>::infinity() : std::numeric_limits<float>::max();
+#else
+        out = NumericTraits<float>::infinity();
+#endif
     }
     else if (abs > 0x3FFU)
     {
diff --git a/libuavcan/src/transport/uc_can_io.cpp b/libuavcan/src/transport/uc_can_io.cpp
index 4df84247d2..e3cc4570cd 100644
--- a/libuavcan/src/transport/uc_can_io.cpp
+++ b/libuavcan/src/transport/uc_can_io.cpp
@@ -6,7 +6,6 @@
 #include <uavcan/transport/can_io.hpp>
 #include <uavcan/debug.hpp>
 #include <cassert>
-#include <limits>
 
 namespace uavcan
 {
@@ -100,7 +99,7 @@ CanTxQueue::~CanTxQueue()
 
 void CanTxQueue::registerRejectedFrame()
 {
-    if (rejected_frames_cnt_ < std::numeric_limits<uint32_t>::max())
+    if (rejected_frames_cnt_ < NumericTraits<uint32_t>::max())
     {
         rejected_frames_cnt_++;
     }
diff --git a/libuavcan/test/marshal/float_spec.cpp b/libuavcan/test/marshal/float_spec.cpp
index a6991c3afc..37dcd74f6a 100644
--- a/libuavcan/test/marshal/float_spec.cpp
+++ b/libuavcan/test/marshal/float_spec.cpp
@@ -3,6 +3,7 @@
  */
 
 #include <gtest/gtest.h>
+#include <limits>
 #include <uavcan/marshal/types.hpp>
 #include <uavcan/transport/transfer_buffer.hpp>
 
diff --git a/libuavcan/test/marshal/scalar_codec.cpp b/libuavcan/test/marshal/scalar_codec.cpp
index 1525ce5281..ee816fdea4 100644
--- a/libuavcan/test/marshal/scalar_codec.cpp
+++ b/libuavcan/test/marshal/scalar_codec.cpp
@@ -3,6 +3,7 @@
  */
 
 #include <gtest/gtest.h>
+#include <limits>
 #include <uavcan/marshal/scalar_codec.hpp>
 #include <uavcan/transport/transfer_buffer.hpp>